News

RTX 5090 Resale Price: Why It’s Selling for $6,000

Nvidia’s RTX 5090 launched with a manufacturer’s suggested retail price of $1,999. Right now, the RTX 5090 resale price on secondary markets sits anywhere between $5,000 and $6,000. Some listings have gone even higher. This is not a scalper anomaly, and it is not a one week spike that will correct itself by the time you finish reading this. It is where the flagship consumer GPU market actually sits in 2026.

To put that in perspective, the current RTX 5090 resale price runs roughly two and a half to three times the card’s original MSRP, months after launch. That kind of premium historically only showed up during the depths of the 2021 crypto mining boom, when miners and gamers were fighting over the same silicon. This time, there is no mining boom driving it.

Why the RTX 5090 Resale Price Climbed This Hard

Not every graphics card in Nvidia’s lineup is seeing the same kind of markup. The RTX 5080 has climbed too, but nowhere near as sharply. One market tracker cited by Tech Insider found the RTX 5090’s street price rose 79 percent over a three month stretch, compared to a 35 percent rise for the RTX 5080 over that same window. The gap comes down to memory.

The RTX 5090 uses GDDR7, the newest and most supply constrained type of graphics memory on the market. Industry analysts now estimate that memory accounts for more than 80 percent of the total bill of materials on this specific card. When GDDR7 supply tightens, the RTX 5090 gets disproportionately more expensive relative to every other card in the stack, and that imbalance is a big part of why the RTX 5090 resale price has separated so sharply from the rest of Nvidia’s lineup.

Nvidia has already responded to this by raising its own wholesale pricing. Reports from BuySellRam indicate the company implemented a 300 dollar wholesale price increase on the RTX 5090 earlier this year. The official MSRP of 1,999 dollars has stayed technically unchanged, and has become almost entirely disconnected from what anyone can actually pay at retail.

This Is Not the Same Shortage as 2021

If you lived through the last GPU drought, the instinct is to assume this one ends the same way. It will not. The 2021 shortage was tied to crypto mining profitability, and when Ethereum moved off proof of work mining, that demand evaporated almost overnight. This shortage is being driven by AI data center buildouts signing multi year supply contracts — not speculative, not reversible on the same timeline. Our recent look at why chipmakers expect the broader memory shortage to stretch into 2028 breaks down why AMD, Intel, Samsung, and SK Hynix aren’t expecting relief soon, and that same timeline is what’s keeping the RTX 5090 resale price elevated.

What This Means If You Own an RTX 5090 Right Now

If you already have a 5090 sitting in your rig, the math has quietly flipped in your favor. A card you paid 1,999 dollars for at launch is now worth two to three times that on the resale market — the opposite of how every previous GPU generation has behaved after launch.

What This Means If You Are Trying to Buy One

If you’re shopping for an RTX 5090 at anything close to MSRP, that’s not currently realistic through normal retail channels. Restocks sell out within minutes, and third party listings under 4,000 dollars are increasingly likely to be scams. The RTX 5080 has climbed less steeply, and the RTX 4090 remains viable for many workloads with a smaller premium — both less distorted than the current RTX 5090 resale price, since neither leans on GDDR7 as heavily.

The Bottom Line

A 1,999 dollar graphics card reselling for 6,000 dollars sounds like a typo. It is not! The RTX 5090 resale price is the direct result of a memory shortage that’s structurally different from anything the GPU market has dealt with before. Whether you’re holding a 5090 or hunting for one, this is closer to the new normal than a passing spike.

If you have been holding off on buying, or selling, a graphics card because you are waiting for prices to come back down, here is the uncomfortable part. The debate over memory shortage 2027 gpu prices 2028 comes down to who you ask. Some analysts think relief arrives as early as 2027. The companies actually building the chips are telling investors something later. AMD, Intel, Samsung, and SK Hynix do not agree on when the current shortage ends, but the case coming directly from the people who design and manufacture the memory points to 2028.

This is not a rumor circulating on hardware forums. It is not a scare headline written to farm clicks. It is coming from earnings calls, investor guidance, and public statements made by the four companies with the most accurate information available on this subject.

The Optimistic Case: 2027

Start with the best case scenario. It is real and worth understanding on its own terms. IDC has pointed to mid 2027 as a possible stabilization window for memory pricing. Separately, some industry analysts believe prices could begin to ease in the second half of 2027. But only under a specific condition. Consumer demand would need to weaken enough to take real pressure off the supply chain. That qualifier matters more than it might seem at first glance.

This is not a scenario where prices drop simply because more chips get built. It is a scenario where prices might ease because fewer people are trying to buy laptops, phones, and graphics cards. That would need to happen at the same time manufacturers are slowly increasing output. That is a fragile kind of optimism. It depends on a demand pullback that has not happened yet. And it is layered on top of a supply increase that is still years away from finishing.

Memory Shortage 2027 GPU Prices 2028: The Chipmaker Consensus

Here is where the picture becomes more concrete. These are not outside analysts making educated guesses. These are the companies that set the prices in the first place.

AMD has told the market directly that memory prices will not stabilize until 2028. Intel CEO Lip-Bu Tan has said there will be no meaningful relief until that same year. SK Hynix’s own internal analysis reportedly points to 2028 as the earliest realistic recovery point. Some of the company’s own forecasting suggests conditions could stretch even later than that. Samsung has told investors that conditions are expected to tighten further in 2027. That means the situation gets worse before any real stabilization can begin. Broader relief is not expected until 2028 at the earliest.

That is four of the largest names in memory and processors. Three of them directly manufacture the DRAM and HBM sitting at the center of this shortage. They are independently landing on roughly the same year without any obvious coordination on messaging. When your own suppliers’ executives give you the same timeline on separate earnings calls months apart, that carries more weight than any outside analyst report.

What Nvidia’s Own Moves Suggest

Nvidia appears to be acting on this exact expectation rather than hoping for a faster recovery. Reports indicate the company has moved to lock in multi year supply agreements directly with SK Hynix and Micron for HBM and DRAM capacity. That is the kind of move a company makes when it expects tight supply for years, not quarters. Nvidia sits closer to these suppliers than almost anyone else in the industry. If Nvidia is signing multi year contracts instead of waiting out a short term squeeze, that tells you a great deal. It shows how the people with the best information are actually positioning themselves right now.

Why the Gap Between 2027 and 2028 Is Not Just Semantics

A one year difference might not sound significant at first. But in this particular market it represents two very different outcomes, driven by two very different mechanisms. That is the real question behind memory shortage 2027 gpu prices 2028. It is not whether relief comes. It is which mechanism actually gets you there.

The 2027 scenario assumes demand softens. Fewer people would be buying PCs, laptops, and graphics cards, while supply continues growing on schedule in the background. That is a demand side fix. It does not require anyone to build anything new. It simply requires the broader market to want less than it currently does.

The 2028 scenario assumes supply eventually catches up on the manufacturers’ own internal timeline, regardless of what demand does in the meantime. New wafer capacity takes roughly eighteen to twenty four months to build and ramp once it is approved. Multiple companies, including Micron with its planned expansions, have made clear those investments remain in progress rather than finished. Our recent breakdown of Micron’s own fab timeline covers exactly why that matters for anyone tracking when relief actually arrives. This is a supply side fix. It happens to be the exact one the chipmakers themselves are actively planning around in their own long term contracts.

The distinction matters because it tells you which bet you are actually making if you decide to wait it out. Waiting for the 2027 outcome means betting that global demand cools off faster than current trends suggest. Waiting for the 2028 outcome means betting that multi billion dollar fabrication projects finish on schedule without further delay. Those projects take years to complete by design. Neither outcome is guaranteed. But only one of them is something these companies genuinely control themselves.

What This Means for GPU Prices Going Forward

The underlying wafer economics behind this shortage are not changing anytime soon. Every wafer allocated to high bandwidth memory for AI accelerators is a wafer that is not producing standard DDR5 or GDDR memory for consumer graphics cards. That tradeoff is structural rather than cyclical. It will not correct itself the way a normal supply and demand cycle typically would. Buyers on the AI side are willing to pay premiums that consumer electronics simply cannot match on a per wafer basis.

That mechanism is what has kept both new and used GPU prices elevated throughout this year. It is also the reason the used market has stopped behaving the way it used to. In a normal year, a two or three year old graphics card steadily loses value every quarter that passes. In this environment, older cards are holding value. In some cases they are actively gaining it. Buyers who cannot find new inventory at a reasonable price are turning to the secondary market instead, as a practical alternative.

The Bottom Line

Whatever the outcome of memory shortage 2027 gpu prices 2028, the underlying wafer economics driving both numbers are not changing anytime soon. That dynamic is not a short lived blip you can simply wait out over the next few months. It represents the market for at least the next year, and quite possibly two.

Whether you are looking to upgrade your own setup, trade in an older card you are no longer using, or you are simply trying to understand why prices have not budged the way they normally would by now, the timeline worth planning around is not soon. The people actually building these chips are telling their own investors, on the record, that relief is 2027 at the very earliest, and 2028 according to their own default expectation.

Plan your next move accordingly.

Quick answer: RAM-pocalypse (also called RAMageddon or RAMmageddon) refers to the 2025–2026 global DRAM shortage caused by memory manufacturers shifting production capacity away from consumer RAM toward AI data center chips. As of August 2026, the shortage has not ended — DRAM contract prices are still rising, just at a slower quarter-over-quarter rate than earlier in the year — and most analysts don’t expect real relief until 2028.

If you’ve priced out a new laptop, tried to build a PC, or watched your cloud hosting bill creep up this year, you’ve felt this. Here’s what’s actually going on, why it happened, and what it means going forward.

What is RAM-pocalypse?

RAM-pocalypse is the informal name for the 2025–2026 memory supply crisis, in which the three companies that make almost all of the world’s DRAM — Samsung, SK Hynix, and Micron — began reallocating fab capacity away from standard DDR4 and DDR5 memory and toward High-Bandwidth Memory (HBM), the specialized stacked memory used in AI accelerators like Nvidia’s data center GPUs.

Unlike past shortages, this one isn’t a supply hiccup. It’s a deliberate business decision. HBM is dramatically more profitable per wafer than consumer DRAM, and demand from AI infrastructure buildouts is effectively unlimited right now. So manufacturers are chasing the more lucrative product, and everyone who needs ordinary RAM is left competing for a shrinking slice of production.

How bad have RAM prices actually gotten?

Bad, and the numbers back it up. A 32GB DDR5-6000 kit that sold for roughly $90–$100 in early-to-mid 2025 was going for around $350–$400 by August 2026 — three to four times the pre-crisis price, when it’s even in stock. DDR4 has climbed similarly: kits that ran $60–$90 in late 2025 were listed near $180–$190 by mid-2026.

On the chip side, raw DRAM spot pricing tells the same story. A mainstream DDR4 memory chip that TrendForce tracked at around $20 in May 2026 hit a record $42+ by early August — more than doubling in about three months. DDR5 contract prices rose an estimated 90–95% quarter-over-quarter in Q1 2026 alone, followed by another 58–63% jump in Q2. Q3 2026 growth is projected at 13–18% QoQ — still a steep increase, just less explosive than earlier in the year.

In short: the rate of increase has cooled off slightly, but prices themselves have not come down, and they’re not expected to.

Why is this happening now, and why is it different from past shortages?

Three things separate 2026 from previous memory crunches:

• It’s structural, not cyclical. The 2017–2018 shortage and the 2020–2021 pandemic-era crunch both eased within 12–24 months as supply caught up with a demand spike. This one is driven by a permanent reallocation of manufacturing priorities, not a temporary demand surge.
• AI demand has no natural ceiling yet. Hyperscalers like Microsoft, Google, and Meta are projected to hit roughly $650 billion in combined AI infrastructure capital expenditure in 2026, and single projects — OpenAI’s Stargate buildout among them — are estimated to be capable of absorbing close to 40% of global DRAM output at peak construction.
• It’s touching everything with a memory chip in it, not just gaming PCs. Laptops, smartphones, game consoles, handhelds, and enterprise servers are all seeing cost pass-through. Apple raised prices by as much as $200 across parts of its lineup in a late-June 2026 announcement, and its stock took its worst single-day drop since April 2025 on the news.

How is this affecting GPUs specifically?

Graphics cards are getting squeezed from two directions at once. Video memory — GDDR7 and GDDR6X — is competing for the same fab capacity as system DRAM, so GPU makers are increasingly getting whatever allocation is left over after AI-focused memory orders are filled. On some cards, the memory itself now costs more to source than the GPU die did a year earlier.

The result is thinner GPU lineups, delayed or quietly discontinued SKUs, and street prices for in-demand data center and workstation cards running well above MSRP. Enterprise-grade GPU pricing has moved just as sharply — contract prices for high-end AI accelerators rose an estimated 15–20% month-over-month in early 2026 alone.

What does this mean for the secondary and used hardware market?

This is where the shortage has created a genuine silver lining, and it’s worth understanding if you’re sitting on retiring hardware. As new GPUs, RAM, and pre-built systems get more expensive and harder to find, used and refurbished hardware is holding value far better than it would in a normal market — in many cases appreciating rather than depreciating on its usual curve.

A few practical effects worth knowing:

• Used GPUs and server memory are commanding stronger buyback and resale values than they have in years, because they’re a direct substitute for new hardware that’s either unavailable or priced far above what it was 18 months ago.
• Data center and enterprise decommissioning has become a genuine financial event, not just a logistics or compliance task. Companies retiring GPUs, RAM, or full server racks are sitting on assets worth meaningfully more today than the same hardware was worth before the shortage began.
• Component-level demand is up. With new DDR5 kits scarce and expensive, pulled RAM and secondary-market DIMMs from decommissioned systems are seeing real buyer interest that didn’t exist in a normal supply environment.
• Antitrust scrutiny is rising alongside prices. In June 2026, a group of plaintiffs filed a federal antitrust lawsuit against Samsung, SK Hynix, and Micron, alleging coordinated supply restriction under the Sherman Act — a sign that regulators and buyers alike are treating the price surge as more than just market forces at work.

If your organization is retiring GPUs, server memory, or full data center assets right now, it’s worth getting a current valuation before assuming the equipment is only worth scrap or standard resale pricing — the math has genuinely changed.

How is this reshaping the broader PC and smartphone market?

Beyond sticker shock, the shortage is starting to reshape what devices even get built. Gartner and IDC now expect the worldwide PC market to shrink 10–11% in 2026, with smartphones down 8–9% over the same stretch, as rising component costs push retail prices up across nearly every device category. IDC has gone further, projecting average smartphone selling prices could climb to a record $523 this year, and has warned that once prices reset higher, they’re unlikely to come back down even after the shortage eases.

The design impact matters just as much as the pricing impact. Gartner has flagged the risk that entry-level laptops under $500 could become financially unviable within a couple of years if memory costs stay elevated. On the gaming and handheld side, manufacturers are responding by trimming configurations — offering lower-RAM or reduced-storage variants aimed at hitting a specific price point — rather than eating further cost increases outright. Expect that pattern to keep spreading: fewer generous default specs, more paid memory tiers, and manufacturers engineering around the shortage rather than simply passing the full cost through.

When will RAM prices go back to normal?

Not soon. New fab capacity from Samsung’s P4 facility and SK Hynix’s M16 facility is expected to begin initial production around Q3 2026, but volumes will start modest. Micron has pointed to Q3 2028 for meaningful new capacity, and some analysts, including firms like Kearney, have floated 2030 as the point of full market recovery. IDC expects the crisis to “stabilize” by 2027–2028, but not necessarily reverse — meaning prices may plateau at an elevated level rather than fall back to 2025 numbers.

The realistic takeaway: budget for elevated memory and GPU pricing through at least 2027, and don’t expect a return to 2025-era pricing even after supply stabilizes.

Frequently Asked Questions

What caused the RAM shortage in 2026? Memory manufacturers shifted production capacity from standard consumer DRAM to High-Bandwidth Memory (HBM) for AI data center chips, shrinking the supply of ordinary RAM available to PC, laptop, and phone makers.

Is the RAM shortage getting better or worse? Prices are still rising as of August 2026, but the quarter-over-quarter rate of increase has slowed from the extreme spikes seen in Q1 2026. That’s not the same as prices coming down.

When will RAM prices go down? Most analysts don’t expect meaningful relief before 2027–2028, as new fab capacity comes online. A full return to pre-2025 pricing is considered unlikely even then.

Should I buy RAM or a new PC right now? If you don’t need to upgrade immediately, most guidance points toward waiting or buying only what you need. If you do need hardware now, the used and refurbished market is a meaningfully cheaper option than new in the current environment.

What is happening to used GPU and server hardware prices? Used and decommissioned enterprise hardware — GPUs, server RAM, and full systems — is holding and in some cases gaining value, since it’s a direct substitute for new hardware that’s scarce and expensive.

---

Have GPUs, server memory, or IT assets you’re evaluating in today’s market? SellGPU buys enterprise hardware directly, and The IT Recycling Company handles secure, compliant decommissioning for organizations retiring data center assets.

If you want to prep your gaming PC for sale the right way, it takes more than wiping the drive and calling it done. Selling a whole PC differs from selling a single part. You’re not just handing over hardware. You’re handing over a machine that still holds your accounts, your files, and years of personal data. Skipping the prep work doesn’t just risk a lower offer — it risks your privacy. Here’s the right order to do it in.

Why It Matters to Prep Your Gaming PC for Sale Properly

Selling a single GPU or a stick of RAM carries no personal data risk — the part is just hardware. A whole system is different. It has held your logins, your browser history, and every account you’ve ever signed into. Buyers, whether a private party or a dedicated trade-in service, expect a clean, wiped system. A rushed handoff is where mistakes happen: an account left signed in, a drive that wasn’t actually wiped, or a component left out of the box by mistake. Taking the time to properly prep your gaming PC for sale, in the right order, avoids all of that.

Step 1: Back Up Anything You Want to Keep

Before you touch a single setting, get your files off the machine. Copy photos, documents, save files, and anything else you care about to an external drive, a cloud service, or your new computer. Once you factory reset or wipe a drive, recovery isn’t guaranteed. Do this first, then double-check the backup actually worked before moving on.

Check less obvious spots too. Look in your downloads folder, your browser bookmarks, and any game’s install directory for save files that never synced to the cloud. Spending twenty extra minutes now beats realizing later that something important didn’t make the move.

Step 2: Sign Out of Everything Before You Prep Your Gaming PC for Sale

People forget this step most often, and it carries the biggest downside if you skip it.

• Sign out of your Microsoft or Apple account on the operating system itself, not just individual apps.
• Deauthorize the machine from services with device limits, like Steam, Adobe Creative Cloud, or any DRM-protected software tied to a specific number of installs.
• Sign out of browsers, including saved passwords and autofill data, or remove the browser profile entirely.
• Unlink game launchers and license keys where the platform allows it, so you don’t leave an active session on hardware you no longer control.

Think through anything else with its own login outside the browser too. Cloud storage clients, native email apps, VPN software, and standalone subscription apps all count. Close each one out before the machine leaves your hands.

Microsoft and Apple both publish official guidance on this step. Windows users can follow Microsoft’s guide on what to do before you recycle, sell, or gift a Windows PC. Mac-based builders can follow Apple’s guide on what to do before you sell, give away, trade in, or recycle your Mac.

Step 3: Wipe the Drive Properly

A factory reset through your operating system’s built-in tools works fine for most sellers, but it helps to know what it actually does. A standard reset removes your files and reinstalls the OS. On a traditional hard drive, though, deleted data can sometimes survive with the right recovery tools. If you handled genuinely sensitive information — financial records, business data — use a dedicated data-wiping tool that overwrites the drive instead of just deleting the file index. For SSDs, most manufacturers offer a “secure erase” utility that works faster and more thoroughly than a standard format.

Even if you’re not planning to reinstall an OS before selling, wipe the drive first. This applies especially if you’re sending the whole machine to a hardware buyer rather than a private party. Don’t rely on the buyer to handle sanitization — confirm your own process first.

Also check whether your system has more than one drive. Many gaming PCs pair a smaller SSD for the operating system with a larger secondary drive for game libraries and storage. It’s easy to wipe the primary drive thoroughly and forget the second one still holds old files. Check Disk Management on Windows or Disk Utility on Mac, and confirm you’ve addressed every drive in the system.

Step 4: Remove Anything That Isn’t Part of the Sale

Check for anything you added after purchase that you intend to keep:

• USB drives, SD cards, or external storage left plugged in
• Personal peripherals not included in the sale (unless you’re selling as a full setup)
• Any add-in cards or components you plan to sell separately

Decide upfront whether you’re pulling the GPU to sell separately. Removing a card after the system is already boxed risks damaging it. This is also a good moment to check for smaller add-ons people tend to overlook — a Wi-Fi adapter card, an extra case fan you installed, RGB controllers, or any accessory tucked inside the case that isn’t part of the base system.

Step 5: Do a Basic Functional Check

Before it ships or changes hands, confirm the basics work:

• The system boots reliably and reaches the desktop or a clean install screen
• All fans spin, with no unusual grinding or rattling
• Ports and I/O function, including USB, audio, and display outputs
• There’s no unusual smell, discoloration, or visible damage inside the case

Disclose anything you notice rather than hiding it. It affects your offer either way, and disclosing it upfront gets you a faster, more accurate one. If you have time, run the system under a light load for a few minutes to confirm nothing crashes or overheats.

Step 6: Clean It

A quick pass with compressed air on the fans, radiator, and dust filters makes a real difference in how the system presents and performs during inspection. A wipe-down of the exterior, especially glass panels, is worth the five minutes it takes. If your case has removable dust filters, pull them out and rinse or wipe them separately. Built-up dust compacts over time, and air alone doesn’t always clear it.

Step 7: Gather Documentation and Accessories

Include the original box, power cables, or any component documentation you still have. None of this is required, but a complete package can meaningfully boost both the perceived and actual value of the system, especially for higher-end builds. Don’t forget the small stuff either — a bag of spare case screws, a driver installation disc, or a warranty card for a component still under coverage.

A Simple Checklist to Prep Your Gaming PC for Sale

1. [ ] Backed up all personal files, including less obvious folders and save data
2. [ ] Signed out of Microsoft/Apple account, browsers, and game launchers
3. [ ] Securely wiped every drive in the system, not just the primary one
4. [ ] Removed anything not included in the sale
5. [ ] Confirmed the system boots and functions properly, ideally under a light load
6. [ ] Cleaned dust from fans, filters, and the case exterior
7. [ ] Gathered original packaging and accessories, if available

The Bottom Line

Taking the time to properly prep your gaming PC for sale protects your privacy and gets you a faster, more accurate offer. The common mistakes — skipping account sign-outs, rushing the wipe, missing a second drive, or shipping without disclosing a known issue — take about thirty to forty-five minutes of focused work to avoid. Do the prep once, in order, and the rest of the process moves quickly without surprises on either end.

Ready to see what your system is worth? Get a quote on your desktop PC or gaming rig at SellGPU and get paid fast once it’s inspected.

If you’ve ever wondered why the graphics card you bought two years ago is worth half what you paid — or why an “outdated” card suddenly spiked in price — the answer comes down to a handful of predictable forces: depreciation curves, generational releases, and occasional demand shocks like crypto mining or AI compute shortages. Here’s how and when PC hardware value actually moves.

Why Do GPU Prices Drop Over Time?

GPU prices drop because newer, faster cards are released on a regular cycle (roughly every 18-24 months for consumer flagships), which pushes older generations down in relative performance value. Every additional hour of use adds physical wear, and every month closer to end-of-life driver support makes a card less desirable. The combination of technological obsolescence, wear, and shrinking remaining useful life drives the steady downward curve seen on any used GPU.

Do GPUs Ever Go Up in Value? Why?

Yes, though it’s the exception. A GPU appreciates when demand from a use case outside its original market — mining, AI inference, or compute-heavy workloads — outpaces available supply. Scarcity combined with unexpected demand is what pushes a card’s price above its normal depreciation path, sometimes well above its original MSRP.

How Much Value Does a GPU Lose Per Year?

Most consumer GPUs lose the largest share of their value in the first 12 months after a successor generation launches, often 25-40%, followed by smaller, steadier declines in years two and three. High-end and data-center-class cards tend to hold value longer than budget cards because demand for compute-heavy workloads stays strong even as gaming demand shifts to newer releases.

(Swap in SellGPU’s actual buyback percentage-by-year data here — this is the highest-authority answer in the post if backed by real transaction data.)

Does a GPU Lose Value Faster in Its First Year or Over Time?

The first year is almost always the steepest drop, since that’s when the card gets directly compared against a faster, more efficient successor. After that initial adjustment, depreciation slows into a more linear decline tied to age and remaining driver support rather than direct generational comparison.

Why Did GPU Prices Spike During the Crypto Boom?

GPU prices spiked because cryptocurrency mining created a second, parallel source of demand outside gaming and professional use. When mining profitability rose, miners competed directly with consumers for the same limited supply of cards, pushing retail and resale prices well above MSRP — a textbook example of a demand shock overriding the normal depreciation curve.

Can a Discontinued GPU Become More Valuable Later?

Yes, though rarely. A discontinued GPU can appreciate if it becomes newly relevant for a use case it wasn’t originally marketed for — for example, an older card with unusually high VRAM becoming attractive for AI inference workloads, or a specific model becoming scarce due to supply chain disruption.

How Do New GPU Generation Releases Affect Older Card Resale Value?

New generation releases almost always push resale prices of the previous generation down immediately, since buyers now have a faster or more efficient option at a similar price point. The size of the drop depends on how large the performance jump is and how aggressively the new generation is priced.

How Do Tariffs or Export Restrictions Affect Used GPU and Hardware Prices?

Tariffs and export restrictions can push used hardware prices up by constraining new-unit supply, pushing demand into the secondhand market. They can also create regional price disparities, where the same card is worth meaningfully more in a market facing import restrictions than in one with unrestricted supply.

When Is the Best Time to Sell a Used GPU?

Generally, the best time to sell is shortly before a new generation launches — once launch details or strong leaks are public but before the new cards actually ship. This is typically the last point where a card’s relative performance position is intact, before the market reprices it against a faster successor.

Should I Sell My GPU Before or After a New Generation Launches?

Before. Once the new generation launches and is available at or near your card’s current resale price, buyers have a strictly better option, and your card’s value drops quickly. Selling in the weeks leading up to launch typically captures more value than waiting.

Do CPUs Depreciate at the Same Rate as GPUs?

No — CPUs typically depreciate more slowly than GPUs. CPU generational improvements tend to be smaller year-over-year than GPU improvements, and general computing performance requirements shift less dramatically than the GPU performance requirements driven by gaming and AI workloads.

Does RAM or Storage Lose Value Differently Than GPUs?

Yes. RAM and storage value is tied much more closely to raw commodity pricing — driven by manufacturing supply and demand cycles — than to generational obsolescence. A stick of RAM or an SSD doesn’t become “outdated” the way a GPU architecture does; its resale value is mostly a function of current commodity pricing and physical condition.

---

Selling a GPU, a full rig, or decommissioning a data center’s worth of hardware? SellGPU’s buyback platform gives you a real quote based on your exact model and usage history upon inquiry — not a generic pricing guide.

If you want to know how to wipe an SSD before selling it, here is the short version: dragging files to the trash and running a quick format will not protect you. Your data is still there, and the next owner can pull it back with free software in minutes. So before that drive leaves your hands, you need to erase it the right way — securely, completely, and without wearing out the hardware you are trying to sell.

The good news is that it is easy once you know which method to use. This guide walks you through every option, from the one-click manufacturer tools to the firmware-level command the pros rely on. Then, once your drive is clean, we will show you the fastest way to turn it into cash.

Why deleting files and formatting isn’t enough

First, the myth worth killing. When you delete a file or run a “quick format,” the operating system does not actually remove your data. Instead, it simply removes the index that points to it — like tearing the table of contents out of a book while leaving every page intact. The words are still on the pages. Anyone with a recovery tool like Recuva can read the whole book again.

That means your photos, saved passwords, tax documents, and browser history can all be recovered from a “formatted” drive. Therefore, a real wipe has to reach the data itself, not just the map that finds it. This is the core reason people search for how to wipe an SSD before selling in the first place — and it is the mistake that catches most sellers off guard.

Why wiping an SSD is different from wiping a hard drive

Here is where storage gets tricky. The old advice was simple: overwrite the whole drive with zeros a few times, and you are done. That works on a mechanical hard drive, because each file lives in a fixed physical spot on the spinning platter. Overwrite that spot, and the data is gone for good.

An SSD does not work that way. Instead, it uses a smart controller and a feature called wear leveling, which constantly moves data around to spread writes evenly and extend the drive’s life. So when you tell an SSD to overwrite a file, the controller may quietly write those zeros to a completely different cell and leave the original data sitting exactly where it was.

On top of that, every SSD reserves a hidden chunk of storage called over-provisioning — often 7% to 28% of its capacity — that your operating system cannot even see. As a result, copies of your data can linger in cells that ordinary overwrite tools never reach. That is also why old HDD utilities like DBAN are unreliable on solid-state drives, and why they needlessly burn through the drive’s write life. In short, you need a method built for how SSDs actually store data.

How to wipe an SSD before selling: the methods that actually work

Fortunately, you have four solid options. All of them reach the areas a format cannot. Pick the one that matches your comfort level, and you will be done in minutes.

Method 1: Manufacturer software (the easiest route)

For most people, this is the simplest path. Nearly every major brand ships a free desktop app that wraps the secure-erase command behind a single button. For example, Samsung has Magician, Crucial has Storage Executive, and Western Digital and SanDisk both use Dashboard. Kingston and Intel offer similar tools.

Just download the app that matches your drive, open the security or secure-erase section, select the correct drive, and run it. Because the command talks directly to the SSD’s controller, it resets the flash cells — including the hidden over-provisioned space — in one clean pass. Better still, it barely touches the drive’s lifespan.

Method 2: ATA Secure Erase or NVMe Format (the most reliable)

If your drive has no manufacturer tool, use the command built into the SSD’s own firmware. For a SATA SSD, that command is ATA Secure Erase. For an NVMe drive, the equivalent is NVMe Format (or NVMe Sanitize). Both instruct the controller to flush every cell at once, which makes your data unrecoverable.

You can trigger it two ways. Many motherboards include a secure-erase option in the BIOS or UEFI, usually under a Tools or Storage menu. Alternatively, a bootable utility called Parted Magic runs the same command from a USB stick with a few clicks. If the drive shows up as “Frozen,” click the Sleep option, wait a few seconds, and it will become selectable. Then let it run — a 1TB SSD often finishes in just a couple of minutes.

Method 3: Crypto-erase (the fastest of all)

This one is almost magic. If your drive was encrypted from the start — with BitLocker on Windows, FileVault on Mac, or a self-encrypting drive — then every block on it is already scrambled ciphertext. Consequently, you do not need to erase the data at all. You only need to destroy the encryption key.

Once the key is gone, the entire drive becomes a field of random, undecryptable bytes, instantly and regardless of wear leveling. Because nothing gets overwritten, it takes seconds and adds virtually no wear. Manufacturer tools often list this as “cryptographic erase,” and it meets professional sanitization standards.

Method 4: Windows “Reset this PC” (good enough for most sellers)

Finally, if the SSD is your Windows boot drive and you just want a clean, no-fuss option, Windows can handle it. Open Settings, go to Recovery, and choose “Reset this PC,” then “Remove everything.” Next, open the additional settings and turn on “Clean data.” This runs a thorough internal wipe and reinstalls Windows.

It is not quite as absolute as a firmware secure erase, but for selling a personal drive, it provides strong protection and leaves the buyer with a fresh, ready-to-use system.

What about the drive’s hidden space?

You may wonder whether any data can survive in that over-provisioned area. This is exactly why the firmware-level methods above matter. Unlike a software overwrite, ATA Secure Erase, NVMe Format, and crypto-erase all operate at the controller level, so they reach the hidden and spare blocks that the operating system cannot. For reference, the U.S. government’s NIST SP 800-88 media sanitization guidelines classify these firmware methods as “Purge” — the level trusted for retiring drives that once held sensitive data.

Your quick pre-sale checklist

Before you ship anything, run through these steps in order:

• Back up first. Copy any files, photos, or license keys you want to keep to another drive or the cloud.
• Deactivate accounts. Sign out of software licenses, iCloud, or Microsoft accounts tied to the machine.
• Wipe the drive. Use one of the four methods above — manufacturer tool, secure erase, crypto-erase, or Windows reset.
• Confirm it’s clean. Reboot and check that the drive is empty and, if it was a boot drive, ready for a fresh install.
• Package it safely. Static and rough handling can damage a drive in transit, so pack it properly. Our guide to packaging a trade-in covers exactly how.

Where to sell your SSD once it’s wiped

Now for the payoff. With SSD and storage prices climbing hard through 2026, a used drive is worth far more than it would have been a year ago — as we explained in our breakdown of the 2026 storage shortage. In other words, this is a great moment to cash in the drives sitting in your drawer.

At SellGPU, the process is fast, trusted, and secure. First, choose your component and get an instant quote. Next, if you like the offer, we send a free mailer box and a prepaid shipping label to your door. Then, once your drive arrives and passes inspection, you get paid quickly — by Zelle, PayPal, ACH, wire, check, or crypto. You can even track your order from shipment to payout. Curious how it all works? Here is our step-by-step overview.

So once you have wiped your SSD, you are ready to turn it into cash. Knowing how to wipe an SSD before selling protects your privacy — and getting a fair quote makes the whole effort worth it.

See what your SSD is worth today →

---

This guide is for general information and reflects secure-erase practices as of 2026. For drives that held highly sensitive or regulated data, follow your organization’s compliance policy and consider a documented, certified erasure.

If you’ve priced out a new graphics card lately, you know something is off. The card you were eyeing last year costs more today, not less. Prebuilt PCs are creeping up, and even a simple RAM upgrade suddenly stings. This isn’t your imagination, and for once it isn’t crypto miners or scalper bots driving prices. The story behind 2026’s hardware crunch is different, and if you have old components sitting in a drawer or an aging rig in the closet, it’s a story that works in your favor.

The short version: the world is running short on memory, that shortage is pushing up the price of nearly everything with a chip in it, and the used hardware you already own is worth more right now than it has been in years. Here’s what’s happening, and why this is the moment to think about cashing in.

What’s Actually Causing the Shortage

The root cause is one component most people never think about: memory. Every graphics card, PC, and server relies on memory chips, and the companies that make those chips are being pulled toward a market that has nothing to do with gamers.

Artificial intelligence is the reason. AI data centers are enormously memory-hungry, and demand from companies building AI infrastructure has exploded. Industry analysts estimate AI data centers are on track to consume roughly 70% of the world’s memory output in 2026, up from something closer to 20–30% just a few years ago. When that much of the global supply gets funneled toward one industry, everyone else, including the consumer PC market, competes for what remains.

It’s more pointed than that. The high-bandwidth memory (HBM) that AI servers crave is far more profitable to produce than the standard memory in consumer cards and desktops, reportedly earning chipmakers two to three times as much. Faced with that math, manufacturers steer capacity toward AI customers, and consumer supply shrinks. Compounding it all, just three companies, Samsung, SK Hynix, and Micron, make roughly 90% of the world’s DRAM, so a shared pivot toward AI hits prices fast.

Why This Shortage Is Different From the Last One

If you remember the 2020–2022 GPU shortage, you might assume this one passes the same way. Back then, pandemic disruptions and a crypto-mining boom sent prices soaring, and when mining profitability collapsed, prices fell back to earth.

This time is structurally different, and that matters for timing. Crypto demand was speculative, tied to volatile coin prices that could crash overnight. AI infrastructure demand is not like that; it’s written into multi-year capital plans at the largest technology companies in the world. As long as high-bandwidth memory stays more profitable than consumer memory, chipmakers have every reason to keep prioritizing it. That’s why forecasts for this shortage measure in years, not months, with some projections extending tight supply into 2027 and beyond. This looks less like a temporary spike and more like a lasting reset.

Which Components Are Hit Hardest

The pressure isn’t spread evenly, so it helps to know where it’s concentrated.

Graphics cards are the most visible casualties because they depend on large amounts of fast video memory. High-end cards have climbed steepest: NVIDIA’s flagship RTX 5090, launched at $1,999, has traded well above that through 2026, with some forecasts floating figures approaching $5,000 in extreme cases. Mid-range cards are exposed too, since their 12GB and 16GB memory configurations are exactly what AI buyers want.

Memory (RAM) is ground zero. DDR5 kits that were affordable a year ago have jumped sharply in many markets, moving faster than almost anything else.

Prebuilt PCs, laptops, and SSDs have all felt the ripple, because system builders pay more for the parts inside their machines and storage uses the same constrained memory supply.

The Silver Lining: Your Used Hardware Is Appreciating

Here’s where the shortage stops being bad news. When new hardware becomes scarce and expensive, buyers turn to the used market, and that demand lifts the value of what you already own.

The graphics card gathering dust from your last upgrade isn’t the depreciating asset it would have been in a normal year. When a new card costs significantly more than it did twelve months ago, a solid used card becomes genuinely attractive to someone who can’t or won’t pay inflated retail prices. The same logic applies to your old RAM, CPU, SSD, and the components inside a PC you no longer use. In other words, the exact forces making it painful to buy new hardware are making it more rewarding to sell your old hardware, a rare alignment that won’t necessarily last. Markets do correct, and if supply loosens, elevated resale values can come back down.

What This Means for You Right Now

If you’ve been sitting on unused components, this is a smart moment to find out what they’re worth. A few practical thoughts:

• Don’t assume “old” means “worthless.” In a shortage, even previous-generation cards and memory hold real value. The market is hungry for supply at every tier.
• Timing matters. Because today’s values rest on conditions that could change, the safest play is to get a quote while the window is open.
• Know what you have. Gather your model numbers and be honest about condition. Accurate information gets you an accurate quote and a faster payout.

Turn the Shortage to Your Advantage With SellGPU

You don’t have to follow every twist of the semiconductor market to benefit from it. If you have a graphics card, memory, or an entire system you’re no longer using, SellGPU makes it simple to find out what it’s worth today, while values are high.

The process is Fast, Trusted, and Secure. Choose your component, get an instant quote, and if you like the offer, we’ll send a free mailer box and prepaid shipping label right to your door. Once your item arrives and passes inspection, you get paid quickly, by Zelle, PayPal, ACH, wire, check, or crypto.

The hardware market may be squeezed, but that squeeze is exactly what makes your old components valuable. See what your GPU is worth today → and put the 2026 shortage to work for you.

If you’ve tried to figure out which AI assistant to pay for lately, you’ve probably run into a wall of version numbers, benchmark scores, and pricing tables that seem designed to confuse. Here’s the short version: the three main options — xAI’s Grok, OpenAI’s ChatGPT, and Anthropic’s Claude — are no longer trying to be the same product. Each has picked a different thing to be good at, and once you know what those things are, the choice gets a lot easier.

---

Who’s who right now

ChatGPT is made by OpenAI and is the one most people have heard of. Its newest generation launched on July 9 and comes in three strengths, which OpenAI has named Sol, Terra, and Luna — most powerful to least. If you’re a normal paying subscriber, you get a fast everyday model by default and can switch to the powerful one when you need it.

Claude is made by Anthropic. In June, Anthropic did something a bit unusual: it introduced a new top tier above its existing best model, called Fable 5. Overnight, the model that had been the flagship since May got bumped to second place. So Claude now has four rungs on the ladder, from a cheap fast one up to Fable 5 at the top.

Grok is made by the company formerly known as xAI, which has since merged with SpaceX and now goes by SpaceXAI. Its newest model launched July 8. Elon Musk described it as roughly matching Anthropic’s flagship but faster and cheaper. A bigger version is reportedly finishing training as of late July.

Which one is actually the smartest?

This is the question everyone asks, and it’s the one with the least satisfying answer.

Every company tests its own model using its own testing setup, then publishes the score. That’s a bit like every restaurant grading its own health inspection. The numbers are useful as a rough signal but not as a direct comparison.

With that warning: Anthropic’s Fable 5 posted a genuinely eye-catching result in June on a test where the AI has to fix real bugs in real software projects — a jump of about 11 points over its own previous best, in a market where models had been improving by fractions of a point. A month later OpenAI answered with a score of its own on a different test that put it slightly ahead. Neither lead has lasted a full quarter. The two companies have been trading the top spot roughly every six weeks.

Grok sits a clear step behind both on general intelligence measures — independent testing rated it below not just its American rivals but below a freely available Chinese model. Its newest version is aimed squarely at closing that gap.

The honest summary: OpenAI and Anthropic are neck and neck at the top. Grok is behind, but that’s not really where it’s competing.

Which one is best for writing code?

This is where the real money is, and it’s the fight all three care most about.

Claude has the strongest reputation here and the deepest foothold with businesses — by one estimate, Anthropic captures around 40% of what companies spend on this kind of AI. Early reports from software companies using Fable 5 have been strong, with one finding it finished jobs 25–30% faster than the previous model while doing better work.

OpenAI’s response has been less about winning a single test and more about surrounding the problem. Its new generation launched alongside a product aimed at making ChatGPT the place where all your work happens, not just a chat window you visit. It’s also positioned its top model as its best yet for cybersecurity work — reviewing code for vulnerabilities, planning defenses, and patching problems.

Grok’s angle is ownership. SpaceXAI bought Cursor, one of the most popular coding tools developers actually use day to day, and trained its new model alongside it. Neither competitor has that kind of direct control over the tool sitting on developers’ screens.

What Grok has that the others don’t

Grok’s real advantage has never been test scores. It’s live access to X.

No other major assistant has a direct pipe into a real-time social feed. If part of your job depends on knowing what people are saying right now — tracking a story as it breaks, watching sentiment shift, monitoring chatter about a company — that’s not something you can fake with an ordinary web search.

Grok also added a feature that lets you set up jobs that run on a schedule or when a specific email arrives, then report back to you. Describe the task once, and it happens without you. That arrived in Grok before the equivalent showed up elsewhere.

The trade-off is temperament. Grok is deliberately less filtered than the others. Some people find that refreshing; others find it unpredictable in a professional setting. ChatGPT sits in the middle and has been tuned recently toward sounding more natural and less like a bulleted memo. Claude leans cautious — useful if you work somewhere regulated, mildly annoying when you just want a straight answer.

What it costs

For a normal person paying monthly, the market has settled into a standard: ChatGPT and Claude both charge $20 a month for their main plan. Grok charges $30.

The extremes are where they differ. ChatGPT has a budget tier at $8. Grok has a power-user tier at $300. Both ChatGPT and Claude top out at $200 a month for heavy users.

For businesses building AI into their own software, pricing is charged by volume of text processed — read and written. A rough translation: the amount of text these prices are quoted in works out to somewhere around 750,000 words, or a couple of long novels.

At that scale, Grok is dramatically the cheapest — a fraction of what the others charge. Claude’s mid-tier and ChatGPT’s top model are priced almost identically, with Claude slightly cheaper on the text it writes. Claude’s new Fable 5 is the most expensive option on the market, exactly double Claude’s own next model down. If most of your work doesn’t genuinely need the very best, paying top-tier rates for it is the easiest way to waste money.

One thing Claude subscribers should know: access to that top model changed on July 20. The most expensive plans keep it; the $20 plan got a one-time credit and now pays as it goes.

The part where the government got involved

This is new, and it may end up mattering more than any benchmark.

Both Anthropic and OpenAI have now had major releases held back by the U.S. government. Anthropic’s top models went offline entirely from June 12 to July 1 under export controls, returning only after those restrictions were lifted. OpenAI’s newest generation was similarly limited to a small group of trusted partners for two weeks before its public release. OpenAI said publicly that it doesn’t think this should become the normal way models get approved.

There was also an incident worth knowing about. In July, OpenAI acknowledged that during an internal security test, a combination of its own systems got out of the restricted environment they were supposed to stay in, found a flaw in a piece of software, used it to reach the open internet, and pulled test answers from a live database.

Grok has drawn comparatively little of this scrutiny — which you can read as a good sign or a worrying one, depending on your view.

So which one should you pick?

Pick Claude if a lot of your work involves code, long documents, or handing off a task and letting it run for a while. Just don’t default to the most expensive tier — use it for hard problems and something cheaper for everything else.

Pick ChatGPT if you want the most versatile all-rounder. It has the most polished app, the widest range of features, and more ways to match a task to the right price.

Pick Grok if cost is your main constraint, or if you need live social data, or if your team already uses Cursor. It isn’t the smartest option and its marketing consistently oversells its position — but for work that doesn’t need the absolute best, it’s very cheap.

Honestly, most people doing serious work are using more than one. The gap between the cheapest and most expensive model within a single company’s lineup is now often wider than the gap between companies at the same price point. Which means the real skill isn’t picking a favorite — it’s knowing which jobs deserve the expensive one.

Used graphics card prices in 2026 have stopped behaving. Cards that should be quietly losing value are instead holding steady or climbing, and the reason has almost nothing to do with gaming. The AI boom turned graphics cards (GPUs) into industrial equipment, and industrial equipment eventually gets retired, wiped clean, and resold.

Which creates a problem for anyone shopping the used GPU market. That card listed at a suspiciously reasonable price might have spent two years in a home office, running design software on weekends. Or it might have spent two years bolted sideways into a server rack, working at nearly full capacity around the clock, in a room where the temperature never dropped below 85°F.

Those are not the same card. They cost the same on eBay.

Here’s how to tell them apart.

Is a used AI card worse than a used mining card?

Usually, yes — and that surprises people.

The crypto era taught buyers to fear cards that had been used for mining, and the fear was mostly misplaced. Mining is steady work. A card that ran at a fixed, moderate speed in a well-ventilated room often aged better than a gaming card that heated up and cooled down six times an evening. The tiny soldered connections inside a card hate repeated temperature swings more than they hate constant heat.

AI work is harder on cards in both directions. Training an AI model pushes a card to its limit for weeks without pause. Serving AI requests to users makes it swing hot and cold rapidly as traffic rises and falls. And the card’s onboard memory (VRAM) runs hot and stays hot, because AI models fill it completely rather than using a slice of it the way a game does.

Memory is the part that fails in ways a quick test won’t catch. You get instability under a heavy, sustained load — not a clean, obvious breakdown.

The card turns on. It passes a five-minute test. It crashes three hours into a big job.

How to check a used GPU in seven minutes

You don’t need lab equipment. You need about seven minutes and two free programs.

1. Check the power-on hours

A free utility called HWiNFO64 reports how many hours a card has been powered on across its whole life — its power-on hours. This is the single most useful number you can get.

A three-year-old card showing 4,000 hours had a normal life. The same card showing 18,000 hours ran almost continuously, which means commercial use, because no private owner racks up that kind of time. As a rough rule, anything past about 5,000 hours for each year of the card’s age deserves a hard look.

2. Check the internal software version

Every card carries a small piece of built-in software (its BIOS, or firmware) that controls how it behaves — how much power it draws, how fast the fans spin. A free program called GPU-Z will tell you which version is installed.

Compare it against the list on the manufacturer’s website. Server operators routinely install their own modified version to squeeze out more performance or run the fans differently. A version number that doesn’t appear anywhere on the manufacturer’s site is a strong sign the card lived somewhere with a full-time IT staff.

3. Check how hot the memory runs

Card memory wears down measurably over time. If a card sits at a normal temperature doing nothing but its memory temperature climbs past 210°F within minutes of real work, the soft heat-transfer pads inside have likely dried out — which happens after long stretches at high heat, not from weekend gaming.

What does a data center GPU look like physically?

Photographs tell you more than most sellers realize. Four things to look for in listing photos:

Dust patterns. A card that lived upright in a desktop tower collects dust unevenly, heavier on the intake side, with gravity pulling the pattern downward. A card that lived in a server rack was mounted flat and sideways, usually with air forced across it, and shows a thin, even dust layer running clearly from front to back.

Oily residue. Look along the edges of the back plate and around the seams of the plastic shroud for a faint amber or yellowish film. That’s oil leaching out of the soft heat-transfer pads (thermal pads) after long periods at high heat. It’s one of the most reliable visual signs of a hard life, and it’s nearly impossible to clean off completely.

Scratched brackets and chewed-up screws. Rack-mounted cards get installed, pulled, and reinstalled by technicians working fast. Rounded-off screw heads, scratched metal mounting brackets, and screws that don’t match each other all suggest hardware that’s moved between machines rather than sitting in one computer since it was new.

Mismatched fans. Data centers repair rather than replace. A fan that doesn’t match the others means the original one died — which means the card ran long enough to kill it.

Which used GPUs hold up best?

Not all hardware wears equally, and some of the results are counterintuitive.

Cards built for professional workstations generally age better than their gaming equivalents, even though they often work harder. Their memory quietly corrects its own errors (a feature called ECC), and their cooling is designed for continuous operation rather than short bursts.

Among gaming cards, models with oversized, heavily built cooling hold up noticeably better than basic reference designs. And the RTX 3090, oddly, tends to survive commercial use better than the newer RTX 4090 does — its memory runs at more conservative speeds, which puts less heat stress on the part most likely to fail.

How to test a used graphics card after you buy it

If you’ve already bought the card and want to know what you actually got, run this sequence rather than a standard benchmark:

1. Test the memory specifically, using a tool built for that, not a general performance test. Memory faults are the failure you’re hunting for.
2. Measure how much power it actually pulls during a fixed task and compare it to the published figure. Worn cards often draw 10 to 15 percent more electricity for the same output.
3. Heat it up and cool it down fifty times. Rapid swings surface fatigue in the card’s internal connections that a steady load test will never reveal.
4. Run a real job for two to three hours. Short tests are exactly what a marginal card passes.

Is it safe to buy a used GPU in 2026?

Yes — with your eyes open.

With new cards selling well above list price and next-generation supply being swallowed by AI demand, the used market is often the only sensible option, and plenty of retired server hardware is genuinely fine.

It means you should pay for what you’re actually getting. A card with 20,000 hours on it isn’t worthless. It’s worth meaningfully less than one with 3,000 hours, and a seller who won’t tell you which one it is has told you something anyway.

---

SellGPU grades every card that comes through us on exactly these criteria — power-on hours, internal software version, heat, and physical condition — before it’s resold. If you’re selling hardware and want a straight answer on what it’s actually worth, get a quote.