Chinese cloud and enterprise customers are locking in long-term server CPU supply agreements with Intel and AMD as prices surge more than 40% over the past year, according to a new report. These deals guarantee purchase volumes for roughly a year without fixing prices — a clear sign that buyers are scrambling to secure capacity in a market where demand is outstripping supply.
What the deals actually look like
The agreements, reported by a Chinese tech outlet, cover both AMD and Intel server processors. Chinese clients — including major cloud providers, telecom operators, and state-owned enterprises — are committing to specific purchase volumes over a 12-month horizon. Crucially, the contracts don't lock in a fixed price, meaning the buyers will pay whatever the market rate is at the time of delivery. That's a gamble, but it's better than the alternative: no guaranteed supply at all.
Prices for server CPUs have jumped over 40% since early 2024, driven by a perfect storm of AI infrastructure buildout, trade restrictions, and limited production capacity at both Intel and AMD. The Chinese market is particularly sensitive to these increases because domestic alternatives from companies like Hygon and Phytium haven't yet matched the performance of x86 chips for high-end workloads.
Why Chinese buyers are locking in now
Several factors are converging. First, the AI arms race has created insatiable demand for compute power, and server CPUs remain essential for the control plane and data preprocessing layers in AI clusters. Second, U.S. export controls have made it harder for Chinese companies to access the most advanced chips, so they're securing whatever they can get. Third, Intel and AMD have been struggling to keep up with orders since the pandemic-era supply chain disruptions, and the situation hasn't fully recovered.
One analyst quoted in the report described the situation as "panic buying" — not for consumer goods, but for the literal engines of the internet. Chinese data center operators can't afford to let their infrastructure go idle, and the cost of downtime far exceeds the premium they're paying for guaranteed supply.
Who wins and who loses
Intel and AMD are the clear winners. These long-term commitments give them revenue visibility and allow them to confidently allocate production capacity. AMD's AMD Instinct GPU line has been grabbing headlines for AI training, but the server CPU business remains a massive cash cow. For Intel, these deals are a lifeline — the company's foundry business is bleeding money, and stable server CPU revenue helps offset those losses.
Chinese buyers lose in the short term. They're paying more for chips that are already marked up due to trade friction. But they also gain predictability in a chaotic market. The alternative — trying to buy on the spot market — would be far more expensive and unreliable.
Smaller Chinese tech firms that aren't big enough to negotiate these long-term contracts are the real losers. Without the volume commitments, they'll face even higher prices and longer lead times, potentially slowing their AI and cloud ambitions.
The bigger picture: a divided silicon world
These deals are a symptom of something larger: the decoupling of global semiconductor supply chains. Chinese companies are increasingly locked into a separate ecosystem where they either buy from Western vendors under restrictive terms or develop domestic alternatives. The long-term contracts are a hedge against future export bans, but they also signal that China's homegrown x86 knockoffs aren't ready for prime time.
At the same time, Intel and AMD are walking a tightrope. They need Chinese revenue — it's a massive market — but they also have to comply with U.S. export controls. The result is a messy compromise where chips get shipped, but under conditions that neither side loves.
A skeptical take: one year isn't long enough
Before we call this a victory lap, let's acknowledge the risk. A one-year commitment without price fixing means buyers are exposed to further price hikes. If the market tightens even more — say, because of new export restrictions or a manufacturing hiccup — those "guaranteed" volumes could become liabilities. The contracts don't guarantee delivery at a reasonable price; they guarantee delivery at whatever the market demands.
There's also the question of whether this is a one-time scramble or a new normal. If Chinese domestic chip production improves faster than expected, these long-term deals could look like overpaying for yesterday's technology. But given the current trajectory, don't bet on it.
What to watch for next
Look for Intel and AMD to announce their own server CPU roadmaps tailored to the Chinese market, possibly with modified designs that comply with export controls while still offering competitive performance. We'll also be watching whether Chinese cloud giants like Alibaba and Tencent increase their investments in Arm-based servers as a way to reduce dependency on x86.
For now, the message is clear: if you need server CPUs in China, you're paying more and you're signing up for the long haul. The era of cheap, abundant compute is over, and both Intel and AMD are making sure they get paid first.
Comments (0)
Loading comments…