GPU Memory Shortage: Why ASUS Stopped Making NVIDIA Graphics Cards [2025]

The GPU market is in crisis mode again. Not the good kind where demand outpaces supply for premium products. The bad kind where manufacturers are making tough choices about what to build and what to abandon.

In early 2025, YouTube channel Hardware Unboxed dropped a story that sent shockwaves through the PC gaming and creator communities. ASUS, one of the world's largest GPU manufacturers, quietly halted production of two high-capacity graphics cards: the NVIDIA RTX 5070 Ti and RTX 5060 Ti, both in 16GB configurations. The reason? Memory supply constraints are so severe that building 16GB GPUs isn't economically viable anymore.

This isn't hyperbole or industry rumor. NVIDIA confirmed it to Engadget, and the implications are significant. We're looking at potential price increases, limited availability, and a restructuring of the GPU market that could last through 2025 and beyond.

Here's what's happening, why it matters, and what you need to know before buying your next graphics card.

TL; DR

• ASUS stopped manufacturing RTX 5070 Ti and 5060 Ti 16GB models due to memory supply constraints making them unprofitable to produce
• Memory prices skyrocketed at end of 2025 driven by AI data center demand, creating a severe shortage affecting GPU manufacturing
• Limited availability will push prices higher for high-capacity GPUs throughout 2025, particularly 16GB and 24GB variants
• NVIDIA claims it continues shipping all SKUs, but third-party manufacturers face real production constraints and profitability pressures
• AI data centers are consuming memory at unprecedented rates, starving traditional GPU manufacturers of critical supply

Estimated data shows a continued rise in GPU prices through 2025, with RTX 5070 Ti prices increasing more sharply due to supply constraints.

The Memory Crisis Explained: Why GPUs Are Disappearing

GPU production isn't just about silicon. It's about memory, power delivery, cooling, PCB design, and dozens of other components that have to work together perfectly. Right now, one of those components is becoming increasingly scarce: memory.

When we talk about GPU memory, we're talking about GDDR6, GDDR6X, or GDDR7 memory chips. These aren't the same as your PC's RAM. They're specialized, high-bandwidth memory designed to work with graphics processors. The demand for these chips has exploded for reasons that have nothing to do with gaming.

AI companies are vacuuming up memory like it's going out of style. Data centers training large language models, building inference clusters, and deploying AI applications are ordering memory in volumes that dwarf traditional consumer demand. When Samsung, SK Hynix, and Micron allocate their memory production, they're following the money. And right now, the money is in data centers, not gaming rigs.

This creates a cascading problem for GPU manufacturers. NVIDIA designs the chips. Board partners like ASUS, MSI, Gigabyte, and PNY source components and assemble the cards. When memory becomes scarce, board partners have choices: pay premium prices for available memory, wait for allocation from suppliers, or stop production of less profitable SKUs.

ASUS chose option three. The RTX 5070 Ti and 5060 Ti 16GB models are expensive to produce because 16GB of GDDR memory is expensive right now. The profit margins don't make sense. So ASUS stopped making them.

NVIDIA's response is diplomatic but telling. "Demand for Ge Force RTX GPUs is strong, and memory supply is constrained. We continue to ship all Ge Force SKUs and are working closely with our suppliers to maximize memory availability." Translation: We're throwing everything at the problem, but we can't magically create more memory.

The company continues to ship all SKUs officially, but that doesn't mean they're easy to find. And it definitely doesn't mean they're affordable. When supply is constrained, prices find equilibrium at whatever level clears the market. Right now, that's higher than it was three months ago.

Estimated data suggests memory supply will catch up with demand by 2026, with potential excess capacity by 2027. Demand may outpace supply if AI investments accelerate.

How We Got Here: The AI Boom's Hidden Cost

The GPU shortage isn't new. We've seen this movie before. In 2021, cryptocurrency mining made GPUs scarce. In 2022 and 2023, AI suddenly became mainstream, creating the first wave of this current crisis. But this 2025 iteration is different and worse in some ways.

Why? Because it's not just GPU supply anymore. The bottleneck has shifted upstream to the memory supply chain itself. You can design better GPUs, but you can't design more memory molecules. Memory production is capital intensive, with multi-year lead times. Samsung and Micron built new fabs, but they're not coming online tomorrow. They're coming online in 2026 or 2027.

Meanwhile, every major cloud company is racing to deploy AI infrastructure. Open AI, Google, Amazon, Microsoft, Meta, and dozens of startups are all building massive GPU clusters. These aren't small deployments. A single training run for a large language model might use thousands of GPUs. And each GPU needs memory. Lots of it.

Consider the math. A single H100 GPU has 80GB of memory. A data center might have thousands of H100s. That's hundreds of terabytes of memory in a single facility. Multiply that across dozens of data centers worldwide, and you're talking about exabytes of memory demand that didn't exist five years ago.

Gaming, creative professionals, and casual users are fighting for scraps. ASUS stopping production of consumer 16GB cards is a signal that the company is making peace with this reality. They're not going to win the memory bidding war against Google and Amazon. They're going to focus on what they can profitably produce and sell.

This market shift has been coming for years. Back in 2023, industry analysts warned that AI would reshape hardware supply chains. Most of the industry ignored them or underestimated the magnitude. Turns out they were right, and the change happened faster than anyone expected.

The Domino Effect: What ASUS's Decision Means for Consumers

When ASUS stops making a product, it's rarely isolated. It's a signal that something deeper is wrong with the market. In this case, ASUS is saying that producing 16GB GPUs isn't economically sustainable at current memory prices and allocation levels.

Other board partners will inevitably follow. MSI, Gigabyte, and PNY are all facing the same memory constraints. They're all doing the same profitability calculations. Some have probably already made the same decision. Others will in the next few weeks.

What does this mean practically? Fewer 16GB GPUs in the market. Period. The RTX 5070 Ti in particular was pitched as a sweet spot: enough VRAM for 4K gaming, creative work, and light AI experimentation without the premium pricing of 24GB cards. Now that option is disappearing.

Price increases are inevitable. Remaining 16GB models from other manufacturers will command premium pricing because they're scarcer. 24GB variants (like the RTX 5070 Ti Super) will see increased demand as users trade up looking for similar performance and VRAM. And 8GB cards will become the only truly affordable option for budget-conscious buyers.

Limited availability also means supply chain adjustments. Usually when one SKU disappears, manufacturers compensate by increasing production of similar SKUs. ASUS will probably make more RTX 5070 (8GB) or 5070 Ti Super (24GB) cards. But these aren't perfect substitutes. The 8GB card is slower. The 24GB card is more expensive. Neither fills the gap quite right.

For gamers, this is annoying but survivable. Most games don't actually need 16GB VRAM yet. For AI practitioners, digital artists, and video editors, it's more serious. These professionals rely on VRAM to handle large datasets, complex scenes, and high-resolution work. Losing the 16GB option forces them to either accept slower cards with less memory or pay substantially more for 24GB variants.

Estimated data shows that AI companies receive approximately 70% of the GDDR memory supply, leaving only 30% for consumer GPU manufacturers. This imbalance highlights the memory supply bottleneck in the GPU market.

Why Memory Supply Is the Real Bottleneck

People often assume GPU shortages are about GPU chips. That's only part of the story. The real constraint is memory, and understanding why requires looking at the memory manufacturing business.

Producing GDDR memory is incredibly complex. It requires specialized equipment, clean rooms, exotic materials, and tremendous capital investment. Samsung and Micron are the primary suppliers for AI and consumer GPU memory. Both companies allocate their production based on long-term contracts and profit margins.

AI companies offer long-term contracts with guaranteed minimum volumes. They're willing to pay premium prices for guaranteed supply. A hyperscaler might contract for 1 million memory units per year at $X per unit. This guarantees Samsung revenue and gives Samsung certainty for capacity planning.

Traditional GPU manufacturers get allocated what's left. They don't have long-term guarantees. If Samsung decides to increase AI customer allocation, consumer GPU manufacturers are the first to lose supply. This is exactly what happened at the end of 2025.

Additionally, memory prices work differently than processor prices. GPUs have relatively stable pricing because NVIDIA sets most of it through MSRP guidance. Memory prices fluctuate based on spot market conditions. When demand exceeds supply, prices spike. This directly impacts board partner profitability.

ASUS probably calculated something like this: "We can make an RTX 5070 Ti 16GB card with

They chose to stop production. It's the economically rational decision, but it's also a sign of just how severe the memory supply problem has become.

Pricing Impact: How Much Are GPUs Really Going to Cost?

Let's talk about money, because that's ultimately what this is about. When supply constraints hit, prices go up. This is basic economics, and we're going to see it across the GPU market in 2025.

For reference, at launch the RTX 5070 Ti had an MSRP of

This pattern will likely continue throughout 2025. NVIDIA's MSRP recommendations don't change, but actual prices do. MSRPs are suggestions that work when supply and demand are balanced. When supply is constrained, prices float to whatever the market will bear.

Meanwhile, 8GB cards haven't seen as dramatic price increases because they're easier to source. Supply and demand are still relatively balanced for lower-capacity GPUs. This creates an interesting market dynamic where the value proposition of different SKUs shifts rapidly.

The RTX 5060 is now a better value than the RTX 5070 Ti when you account for availability and actual prices. It's slower, but you can actually buy it at a reasonable price. This might sound obvious, but it fundamentally changes how consumers evaluate GPU purchases.

Professional users face an even tougher situation. If you need 16GB for your work and can't get it from ASUS anymore, you're either buying from remaining inventory (at inflated prices), waiting for other manufacturers to produce more (which may never happen), or trading up to 24GB cards (which cost significantly more). There's no good option.

Looking forward, prices will probably stabilize once new memory capacity comes online. Samsung and Micron have new fabs starting production in 2026 and 2027. Once those are operational, memory supply will increase and prices will moderate. But that's more than a year away. Until then, expect premium pricing for anything over 8GB.

High-capacity GPU prices are projected to increase by 20-40% through mid-2025 due to memory constraints, while standard GPUs see smaller increases. Estimated data.

The AI Bubble Effect: How Data Centers Are Reshaping Hardware Markets

This entire crisis flows from one source: explosive AI infrastructure investment. Every major tech company has announced massive capital expenditure plans. In 2024, data center and AI infrastructure spending exceeded

That's not just GPUs. That's memory, networking, power systems, cooling, facility construction, everything. But GPUs and memory are the primary bottleneck right now, so that's where the crunch is most visible.

The interesting question is whether this is sustainable. Can the industry continue investing at this pace? Are we in a genuine AI infrastructure boom or an AI bubble that's going to pop?

Historically, when companies make massive synchronized capital investments in similar infrastructure, it often ends badly. Everyone builds too much capacity, prices collapse, and some companies go bankrupt. Think about the telecom bubble of 2000 or the data center bubble of 2010.

But AI infrastructure might be different. The demand for AI compute isn't speculative. Companies actually need to train models, serve inference requests, and build applications. This isn't about theoretical applications that may never materialize.

Still, there's a risk. If one of the major players hits diminishing returns or slows investment, it could create a supply shock in the opposite direction. Memory prices could collapse. GPU prices could crater. This would be good for consumers but devastating for manufacturers who've been building capacity and paying premium prices for components.

ASUS's decision to stop making 16GB cards should be understood in this context. The company is hedging. It's not betting that demand will sustain high prices for premium SKUs. It's betting that memory constraints will persist for long enough to make those cards unprofitable, and that demand will shift to lower-capacity alternatives and lower-margin business segments.

Alternatives: What Builders and Professionals Should Consider

If you're in the market for a GPU in early 2025 and facing limited options, what should you actually do?

For gamers, the situation is relatively straightforward. Most games don't need more than 12GB VRAM. The RTX 5070 (8GB) can handle 4K gaming at high settings in most titles. If you're a competitive multiplayer gamer, 8GB is plenty. If you're targeting maximum quality and highest framerates, stepping up to 12GB or 16GB makes sense, but it costs significantly more right now.

The pragmatic move is to accept the constraints and choose accordingly. Don't overspend trying to future-proof for games that don't exist yet. Buy what meets your needs today, knowing that in 18-24 months, better GPUs at lower prices will be available.

For AI practitioners, the situation is trickier. If you're training models or running inference, VRAM often matters a lot. Less VRAM means smaller batches, slower training, or using less capable models. This can genuinely impact your work.

One option is looking at professional cards. NVIDIA's RTX 6000 and RTX Blackwell professional series have the same or better performance as consumer cards but with enterprise supply chains that haven't been disrupted as severely. They cost more upfront, but availability is better and reliability is higher. If your work depends on GPU access, professional hardware might be a better long-term bet.

Another option is cloud compute. Renting GPU time from AWS, Google Cloud, or Azure gives you access to unlimited GPU capacity without buying hardware. If you only need GPUs occasionally, cloud is cheaper. If you need them constantly, buying your own eventually becomes cheaper, but availability risk disappears.

For digital content creators, the story is similar. 16GB of VRAM is convenient for complex scenes and high-resolution work, but it's not essential. Most rendering workflows can work around limited VRAM by using smart memory management, working in tiles, or using lower-resolution previews during editing.

ASUS's decision to halt 16GB GPU production leads to reduced availability and increased prices for these models, while 24GB and 8GB GPUs see different market dynamics. Estimated data.

Manufacturer Responses: What's Each Company Doing?

NVIDIA's official stance is that it continues shipping all SKUs and is working with suppliers to maximize memory availability. This is technically true but somewhat evasive. NVIDIA doesn't face the same supply constraints as board partners because it doesn't need to source components on the spot market. NVIDIA allocates chips to partners based on demand, but partners have to source memory and other components themselves.

Meanwhile, board partners are making different choices. ASUS stopping 16GB production is notable, but it's unlikely to be alone. MSI has hinted at supply challenges. Gigabyte has mentioned production adjustments. PNY hasn't made public statements, but they're probably managing similar constraints.

The interesting question is whether we'll see more dramatic cuts. Will board partners drop other high-capacity SKUs? Will 12GB models become standard for mid-range cards instead of 8GB? These adjustments could happen gradually or all at once depending on how memory supply evolves.

AMD is in a different position. AMD's GPUs are less popular in consumer markets, so demand and supply are more balanced. AMD faces its own challenges, but it's not dealing with the same production cut decisions. This could actually be a competitive advantage for AMD if it can position its cards as more available alternatives. AMD Radeon 5700 XT variants might become attractive simply because they're purchasable.

Intel's Arc GPUs are still ramping production, so this situation actually works against them. They're trying to build market share, but memory supply constraints are hitting them hard. Intel needs to compete on value and availability, but availability is exactly what's being constrained.

Supply Chain Rebalancing: When Will This End?

The memory supply situation will eventually normalize, but the timeline is uncertain. Several factors could accelerate or delay recovery.

Positive factors: Samsung and SK Hynix have announced new memory fabs. Micron is expanding capacity. These facilities take years to build but once operational, memory supply will increase substantially. Current projects suggest meaningful capacity additions in 2026 and 2027. Within two years, memory supply might catch up with demand. Within three years, there's probably excess capacity.

Negative factors: If AI infrastructure investment continues accelerating, demand could outpace new supply additions. Hyperscalers might lock in longer-term contracts at higher prices, further constraining spot market availability. Geopolitical tensions could disrupt manufacturing or supply chains. A recession could reduce AI spending but also reduce consumer GPU demand, potentially creating oversupply and industry turbulence.

The most likely scenario is gradual normalization through 2025 and 2026. Memory prices probably stay elevated through mid-2025, then gradually moderate as new capacity comes online. GPU pricing follows memory pricing with a lag, so we probably don't see normal pricing until late 2025 or 2026.

For consumers, this means patience is rewarded. If you can wait until late 2025 or 2026, GPU pricing and availability will be significantly better. If you need a GPU today, accept the constraints and make pragmatic choices about capacity and performance.

GPU prices are expected to remain elevated until late 2025, with availability improving significantly by 2026. Estimated data based on current market predictions.

Strategic Implications: What This Reveals About the Market

ASUS's decision to stop making 16GB consumer GPUs reveals something important about how markets actually work. When theory meets reality, when supply gets truly constrained, companies make hard decisions. ASUS could have kept making 16GB cards and absorbed lower margins or passed costs to consumers. Instead, it chose to exit that market segment.

This signals confidence that the constraint is real and lasting. If ASUS thought this was temporary, it would push through and maintain market presence. The fact that it's cutting SKUs suggests the company believes memory supply will be constrained for long enough to justify exiting the category.

It also reveals the hierarchy of priorities in manufacturing. ASUS would rather make margin on lower-capacity cards and professional products than sustain lower-margin consumer premium SKUs. This is a rational business decision that prioritizes profitability over market share.

For the broader industry, it suggests that we're going to see substantial market restructuring. Companies are going to focus on products they can profitably manufacture with current supply constraints. This might mean consolidation around fewer SKUs, focus on value segments, and reduced product diversity.

Consumers should expect this to last. Companies don't make permanent product cuts lightly. When ASUS stops making something, it's usually because it doesn't plan to come back soon. Plan around that reality.

The Broader Context: Hardware Shortages as New Normal

This isn't the first GPU shortage, and it probably won't be the last. Looking at the history of hardware availability, constraints have become more common, not less.

We had GPU shortages in 2021 and 2022 from cryptocurrency mining. We had memory shortages in 2000 and 2018 from various causes. We had component shortages in 2021 from pandemic-related disruptions. The pattern is clear: specialized components have supply curves that don't adjust smoothly when demand spikes.

What's different about this situation is the persistence. Memory supply constraints aren't going away in weeks or months. They're structural problems that will take years to resolve. This is closer to the memory shortage of 2018, which lasted about 18 months.

For businesses that depend on GPU availability, the lesson is to build redundancy and diversification. Don't assume that your preferred hardware will be available. Have fallback options. Consider cloud compute. Explore multiple GPU architectures. Build software that's hardware-agnostic.

For consumers, the lesson is to buy when you need something, not when you want something. Waiting for prices to drop has become a worse strategy because supply constraints are unpredictable. But buying premium SKUs hoping to future-proof yourself is now equally risky because those SKUs might disappear.

The middle path is to match hardware to actual needs and be pragmatic about timing. If you need a GPU, buy one that meets your current requirements at a price that makes sense. Don't overshoot trying to hedge against future needs or overprice for capacity you probably won't use.

What NVIDIA Says vs. What's Actually Happening

NVIDIA's statement deserves closer examination. The company said it "continues to ship all Ge Force SKUs." This is technically accurate but omits important context.

NVIDIA ships reference designs and allocates chips to board partners. But allocation volumes have probably shifted. NVIDIA might be allocating more chips for 8GB variants and fewer for 16GB variants, even if it's technically shipping the full SKU range.

Additionally, NVIDIA doesn't face the same manufacturing constraints as board partners because NVIDIA doesn't manufacture or assemble anything. NVIDIA designs chips and contracts with TSMC for manufacturing. NVIDIA doesn't source memory or PCB components. Those costs and supply risks fall on partners.

So when NVIDIA says it continues shipping all SKUs, it's not addressing the actual problem. The problem isn't that NVIDIA won't ship the chips. The problem is that board partners can't profitably assemble cards from those chips at current material costs.

This is the classic difference between a fabless design company and an integrated manufacturer. NVIDIA can maintain SKU breadth because it doesn't absorb manufacturing costs. Board partners have to make hard choices.

NVIDIA's statement also mentions "working closely with our suppliers to maximize memory availability." This sounds proactive, but it's worth asking what this actually means. Is NVIDIA paying premium prices to secure more memory for board partners? Is it negotiating long-term contracts? Is it just working with suppliers on allocation priorities?

The answer is probably some of each, but we don't have full visibility. What we do know is that it hasn't been enough to prevent ASUS from cutting production of high-capacity cards.

Looking Ahead: Predictions for GPU Markets in 2025 and Beyond

Based on current supply constraints and demand patterns, here's what probably happens over the next 18-24 months.

In the first half of 2025, memory constraints persist. GPU prices remain elevated, especially for high-capacity models. Board partners continue rationalizing their SKU portfolios. We might see more companies dropping 16GB models or consolidating around 8GB and 24GB variants.

Availability gradually improves through mid-2025 as new memory capacity comes online. Memory prices start moderating. GPU pricing follows with a lag. By late 2025, pricing is approaching normal levels, though still elevated compared to 2023.

In 2026, assuming no major supply disruptions, memory supply normalizes. GPU pricing returns to historical norms. Availability improves substantially. Board partners might reintroduce SKUs they dropped in 2025, or they might decide those products aren't necessary anymore.

The wildcard is AI demand. If AI infrastructure investment continues accelerating, demand for GPUs and memory could stay elevated indefinitely. If AI investment moderates or pauses, supply could outpace demand, creating the opposite problem: oversupply and price deflation.

Most likely, we end up with a new equilibrium where AI infrastructure consumes a significant chunk of GPU and memory production, consumer markets get what's left, and supply tensions are chronic rather than acute.

For strategic purposes, assume GPU availability will be constrained through at least mid-2025. Plan accordingly. Don't make purchasing decisions assuming normal market conditions will return quickly.

The Bigger Picture: Hardware as a Bottleneck for Innovation

Zooming out, ASUS's decision to stop making 16GB GPUs is a symptom of a larger challenge: hardware availability is becoming a genuine constraint on innovation.

When researchers want to train models, build AI applications, or push the boundaries of what's possible, they increasingly run into GPU and memory constraints. Not because the algorithms don't exist or because the software isn't ready. They run into constraints because the hardware isn't available or affordable.

This creates a form of innovation inequality. Well-funded companies and well-connected researchers can access GPUs through cloud providers or custom contracts. Researchers at universities or bootstrapped startups run into availability walls.

Over time, this probably favors consolidation toward larger companies and institutions that can secure hardware through preferential allocation or capital resources. Innovation might slow because promising projects can't get the compute they need.

This is a real problem that doesn't have easy solutions. You can't magically increase memory production. You can't force NVIDIA to allocate more chips to consumer markets. You can't prevent hyperscalers from buying all available compute.

What you can do is encourage memory manufacturers to invest in capacity, support open-source alternatives to proprietary solutions, and distribute compute more equitably. But these are long-term structural changes that require sustained effort.

For now, hardware availability will remain a constraint. It's worth understanding that context when evaluating why ASUS made the decisions it made.

FAQ

What exactly is ASUS doing and why?

ASUS has stopped manufacturing the RTX 5070 Ti and RTX 5060 Ti in 16GB configurations. The company made this decision because GDDR memory prices have increased so much that producing these cards would result in negative profit margins. Rather than absorb losses or significantly raise prices, ASUS chose to exit production of these high-capacity SKUs and focus on more profitable products.

Is this temporary or permanent?

Based on ASUS's public statements and industry patterns, this appears to be long-term rather than temporary. The company wouldn't cut SKUs lightly, and memory supply constraints are expected to persist through at least mid-2025. ASUS is probably planning to keep these cards out of production until memory becomes more affordable.

Will other board partners do the same thing?

Very likely. MSI, Gigabyte, PNY, and other board partners face identical supply constraints and identical profit margin calculations. We may already be seeing similar cuts from other companies, or cuts could be announced in coming weeks. The industry-wide trend will probably be consolidation toward lower-capacity SKUs that are easier and more profitable to produce.

How much will GPU prices increase because of this?

Prices have already increased 20-40% for high-capacity GPUs compared to launch MSRP. Further increases are likely if supply remains constrained. 8GB models have seen smaller price increases because they're easier to source. 16GB and 24GB models command significant premiums. The price increases probably continue through mid-2025 as memory supply remains tight.

Should I wait to buy a GPU or buy now?

It depends on your situation. If you need a GPU immediately, buy now but be pragmatic about capacity. Accept that 8GB variants are available and affordable, while 16GB variants are scarce and expensive. If you can wait until late 2025 or 2026, prices and availability will likely be significantly better. Don't try to buy 16GB cards right now unless you absolutely must have that capacity today.

What causes this memory supply crunch?

AI infrastructure demand has exploded, with data centers consuming massive quantities of GDDR memory for training and inference. Meanwhile, traditional GPU manufacturers compete for the same limited memory supply. Memory manufacturers (Samsung, SK Hynix, Micron) prioritize AI customers who offer long-term contracts and premium prices. Consumer GPU manufacturers get allocated what's left. When demand far exceeds allocation, board partners face uneconomical production costs and cut SKUs.

Could this happen again in the future?

Absolutely. Hardware supply chains are inherently vulnerable to demand spikes in specialized components. If another industry suddenly needs massive quantities of a scarce component, similar shortages could occur. GPU memory isn't unique in this regard. We should expect periodic constraints whenever demand outpaces manufacturing capacity for specialized components.

Are there alternatives I should consider?

For gamers, honestly, 8GB is probably enough for most games in 2025. For professionals who genuinely need high VRAM, enterprise cards like the RTX 6000 have better supply chains, or cloud GPU services like AWS or Google Cloud give you unlimited access without worrying about hardware availability. AMD Radeon cards might also be worth considering if they're more available in your region.

When will GPU pricing return to normal?

Memory supply will probably normalize gradually through late 2025 and into 2026 as new manufacturing capacity comes online. GPU pricing typically follows memory pricing with a 2-4 month lag. You might see significant price drops by Q4 2025, with truly normal pricing by mid-2026. That's 12-18 months away, so patience is recommended if you can manage it.

What's NVIDIA's role in this?

NVIDIA designs the GPUs and allocates chips to board partners, but doesn't manufacture anything itself or source components. NVIDIA continues shipping all SKUs to partners who can afford them, but NVIDIA isn't responsible for the manufacturing cost constraints that board partners face. NVIDIA's official statement is accurate but sidesteps the real problem that partners are facing.

Conclusion: The Reality of Hardware Supply in 2025

ASUS stopping production of RTX 5070 Ti and 5060 Ti 16GB cards isn't a random corporate decision. It's a symptom of fundamental supply chain imbalances that will shape hardware markets throughout 2025 and beyond.

Memory has become the bottleneck. Not GPU design, not manufacturing capacity for chips, not cooling or power delivery. Memory. And memory production can't increase overnight. It takes years to build new manufacturing capacity, and that capacity comes online gradually.

Meanwhile, AI companies are consuming memory at unprecedented rates, willing to pay premium prices for guaranteed supply. Consumer GPU manufacturers can't compete with that. ASUS made a rational decision to stop competing and focus on products it can profitably manufacture.

For consumers and businesses, this means several things. First, accept that GPU availability will be constrained through at least mid-2025. Plan accordingly. Second, be pragmatic about specifications. Buy what you need today, not what you think you might need in 2027. Third, consider alternative solutions like cloud compute for tasks where hardware purchase doesn't make economic sense.

Most importantly, understand that this is probably just the beginning. As AI infrastructure investment continues and memory demand stays elevated, we might see more SKU cuts from other manufacturers. The GPU market will gradually consolidate around fewer options optimized for profit rather than breadth of choice.

This isn't necessarily bad. Fewer SKUs mean simpler decisions. Consolidation drives manufacturing efficiency. But it does mean less flexibility and higher prices for the consumer options that remain.

The hardware supply crisis of 2025 is real, it's not going away quickly, and it's time to plan around that reality rather than hoping it disappears. ASUS's decision to stop making certain cards is just the latest signal that the market is adjusting to new constraints. More signals are coming.

Pro tip for staying ahead of supply issues: Monitor official board partner websites and retailer inventory levels rather than relying on MSRP pricing or general industry news. Actual supply constraints show up in product availability before they show up in price changes. When you see a SKU consistently out of stock, that's a signal that the company might be discontinuing it rather than just temporarily unavailable.

Key Takeaways

• ASUS stopped manufacturing RTX 5070 Ti and 5060 Ti 16GB GPUs because memory costs made production unprofitable—a sign of severe supply constraints
• AI data centers are consuming massive quantities of GDDR memory, starving traditional GPU manufacturers of critical supply they need for consumer products
• GPU prices have increased 20-40% for high-capacity models, with further increases likely through mid-2025 as memory constraints persist
• New memory manufacturing capacity won't come online until 2026-2027, meaning supply constraints will remain chronic throughout 2025
• Consumers and businesses should expect limited GPU availability and higher prices through at least mid-2025, requiring pragmatic decisions about capacity needs

Related Articles

• NVIDIA RTX 5070 Ti Discontinued: Memory Crisis Forces Production End [2025]
• RAM Price Hikes & Global Memory Shortage [2025]
• Nintendo Switch 2 RAM Costs, Tariffs & Console Pricing [2025]
• Samsung RAM Price Hikes: What's Behind the AI Memory Crisis [2025]
• NVIDIA RTX 3060 Comeback: Why 2021 GPUs Return to Fight AI Shortage [2025]
• PC Market Downturn: How AI's Memory Crunch Will Impact 2026 [2025]