How AWS and the Hyperscalers Are Cornering Global Hardware Supply

This gives AWS a self-reinforcing advantage: every dollar of capital expenditure can be amortized over a longer period than any enterprise buyer can achieve, and the scale of buying means preferential pricing on every component.

AWS's Trillion-Dollar Ambition

AWS is already on a $169 billion annualized revenue run rate with 36.7% year-over-year growth and a 39.4% operating margin . But Jassy's ambition goes much further. On the Q2 2026 earnings call, he stated that AWS could "very possibly be a trillion-dollar annual revenue business for us in time" — up from his previous estimate of a few hundred billion dollars .

To support that vision, Amazon now expects approximately $220 billion in cash capital expenditure in 2026, up from an earlier ~$200 billion estimate, with the majority going to AI and AWS infrastructure . Yet even at that staggering level, Jassy acknowledged that Amazon will not have enough computing capacity to meet demand in 2026 and possibly through 2027 .

The "RAMpocalypse": Memory Costs Soaring

One of the most significant developments is the explosion in memory costs, which The Register and others have dubbed the "RAMpocalypse." All four major hyperscalers (Microsoft, Amazon, Google, Meta) buy HBM memory from SK Hynix, and HBM accounts for roughly 30-40% of GPU bill-of-materials costs . Hyperscalers' combined purchase commitments total nearly $2 trillion, with a significant portion allocated to memory, giving them outsized leverage to negotiate prices, volumes, and allocation priority with memory makers .

The consequences for everyone else are severe:

• DDR4 prices climbed 158% and DDR5 prices surged 307% since September 2025, according to independent analyst firm TrendForce .
• OVHcloud CEO Octave Klaba warned in August 2026 that server rental prices may rise by up to 87%, with the price OVHcloud pays for memory having risen 6x in the year to June 2026, on track to 9x by September and 12x by 2027 .
• OVHcloud projects RAM prices may rise 250-300% by the end of 2026 compared to September 2025, with pricing not expected to normalize before 2028 .

Because hyperscalers buy at contracted volume prices, the spot-market pain hits smaller providers and enterprise buyers worst. OVHcloud, trying to absorb some of the cost, has limited its average price increase to 9-11% for cloud deployed between 2026 and 2028, but its newer gaming servers will see an 87% jump .

Third-Tier Priority for Enterprise Buyers

The hardware supply chain is now effectively stratified. According to Omdia's forecast as reported by The Register, three tiers have emerged :

• Tier 1: Hyperscaler AI hardware — gets first priority from chipmakers and memory manufacturers.
• Tier 2: Hyperscaler general-purpose hardware — second priority.
• Tier 3: Traditional enterprise buyers — third priority, meaning longer lead times, fewer configuration options, and higher prices.

This structural shift was confirmed by Nutanix CEO Rajiv Ramaswami, who pointed out in May 2026 that the fastest way to access a new server was to rent it from a cloud provider rather than trying to buy it yourself .

Meta Enters the Fray

Adding to the supply pressure, Meta is now actively building a cloud business to sell excess AI compute capacity . At the May 2026 shareholder meeting, CEO Mark Zuckerberg said entering cloud computing was "definitely on the table" . By July, Bloomberg and Reuters reported that Meta was developing plans for a cloud infrastructure business that would sell access to AI computing power and models, setting up a new vector of competition with AWS, Microsoft Azure, and Google Cloud .

If Meta enters cloud infrastructure, it adds another hyperscale buyer into the same fixed pool of global memory, GPU, and server supply — further tightening availability .

What This Means for Enterprise IT

For organizations that still prefer owning their infrastructure or using smaller cloud providers, the downstream consequences are stark:

• Supply starvation for general-purpose servers. Longer lead times, fewer configuration options, and higher prices for anyone buying servers for on-premise data centers .
• Memory cost pass-through. Even if you buy your own servers, memory prices have climbed 158-307% since September 2025 . The cost to refresh or expand on-premise infrastructure has dramatically increased.
• The rental cost squeeze. OVHcloud's 87% price hike warning for gaming servers and 40-59% increases for other recent servers is a leading indicator. Smaller cloud providers cannot absorb memory cost inflation — they must pass it to customers .
• AWS itself can't meet demand. Even customers willing to pay AWS prices face allocation constraints, and AWS can prioritize high-margin, long-term AI workloads over general-purpose compute .
• The migration math flips. With 85% of global IT spend still on-premise and hyperscalers arguing their unit economics make cloud cheaper (server payback in <3 years vs. 5-6 year useful life), enterprise CFOs face pressure to shift workloads to cloud despite rising rental costs — because the alternative may be even worse.

The Bottom Line

The Register's analysis paints a picture of a structural market shift: hyperscalers are using AI-driven demand to lock up multi-year component supply at preferential prices, amortize infrastructure over 30-year horizons, and establish a self-reinforcing cycle where their scale makes them the only buyers who can absorb current component costs. Enterprise buyers who prefer on-premise ownership face rising hardware costs, longer lead times, and memory prices up 300% — while cloud rental prices from smaller providers may rise 87%. Even AWS, spending $220 billion in a single year, admits it cannot build fast enough to meet all demand.