Amazon CEO Jassy warns AI inference remains too expensive as AWS capacity sells out into 2028
Amazon CEO Andy Jassy revealed that Amazon Web Services compute capacity is booked years in advance, with 2026 sold out and commitments stretching into 2028, even as high inference costs continue to slow widespread AI adoption.
Artificial intelligence inference remains "an order of magnitude too expensive" for pervasive enterprise adoption, according to Amazon CEO Andy Jassy in an interview with venture capital firm a16z. Even with current pricing headwinds, demand for cloud infrastructure has surged so rapidly that Amazon Web Services has already sold out its major compute commitments for 2026, booked most of its capacity for 2027, and sold a significant portion of 2028, though AWS still reserves some capacity for on-demand use.[3][5][2][4]
Jassy warned that bottlenecks in data centers, power, and memory supply are unlikely to ease before 2027. These persistent constraints are forcing Amazon to make difficult allocation decisions, splitting scarce resources such as memory between rapidly growing AI workloads and AWS's core non-AI cloud business.[3][1][6]
To address hardware costs and compute shortages, Jassy pointed to Amazon's internal silicon efforts, including its Trainium chips, following what he described as one of the best acquisitions in Amazon history with Annapurna Labs. AWS claims its custom AI accelerators offer roughly 30% to 40% better price-performance than comparable alternatives. While Amazon had previously guided toward roughly $220 billion in capital expenditures for 2026, Jassy noted the company could invest even more without it being a bad use of capital, bolstered by strong demand that has driven Amazon shares up nearly 30% year-to-date.[3][5][6]
Key facts
- Andy Jassy stated that AI inference is 'an order of magnitude too expensive' for widespread enterprise adoption.
- AWS cloud computing capacity is sold out for 2026, largely booked through 2027, and has a good chunk of 2028 already sold, excluding reserved on-demand capacity.
- Jassy indicated that shortages of power, data centers, and memory will not end in 2027.
- AWS claims its in-house AI accelerators provide approximately 30% to 40% better price-performance compared to alternatives.
- Jassy indicated Amazon could invest more than its previously projected $220 billion in capex for 2026 because of intense compute demand.
- Amazon's stock has climbed to its highest level of the year, gaining nearly 30% year-to-date.
Sources · 5 sources
- SB
Shay Boloor@StockSavvyShayPost on X ·
$AMZN CEO Andy Jassy says "a good chunk of '28 is sold out" while AWS is already having to split scarce memory between AI accelerators and its core cloud business. That's now the world's largest cloud provider confirming what $SPCX and $NBIS have been saying all month with SpaceX unable to find spare compute and Nebius saying it could sell out 2027 capacity at today's prices. It's getting harder to see where the AI compute glut bears expected in 2028 is supposed to come from.
Open source - TE
Techmeme@TechmemePost on X ·
Andy Jassy says AWS capacity for 2027 is "largely sold out" and a "good chunk" of 2028 capacity is also sold, though AWS reserves some capacity for on-demand (@catherineperlo1 / The Information) (Visit Techmeme dot com for the link and full context!)
Open source - SE
Serenity@aleabitoredditPost on X ·
So back on your Neoclouds like $NBIS + memory read through... From $AMZN CEO + A16z interview. Q: On memory shortages (eg. $MU / SK Hynix / Samsung): A: "There's only so much memory, and I don't think that that shortage ends in 2027." Q: On AWS capacity: A: "A good chunk of 28 is sold out", "largely sold out through 2027." Q: On capex ROI: A: Amazon had said it would spend about $220 billion in capex for 2026. But Jassy said the company "could invest a lot more and it wouldn't be a bad use of capital" This is probably validating economics of Neoclouds model since this comes from a pretty stingy hyperscaler like Amazon... that only serves free bananas to their employees. Unlike trusting $META / Zuck where you the guy throws all you can eat cruise ship type buffets at every employee. - AWS claims approximately 30-40% better price-performance than comparable AI accelerators. Then went on to say Annapurna was "one of the very best acquisitions we've done in Amazon and the history of the company, if not the best." "We spun out a separate team in Annapurna to build an AI accelerator" Just to give a bump to your AI custom silicon type private companies (since others like $AAPL were likely going on shopping sprees). TLDR: - Not really too much groundbreaking, just incremental validation of what was said previous earnings. - Broader shortages (like memory) won't end in 2027 - Always positive to hear from $AMZN on insatiable demand + ROI given atrocious Neocloud sentiment right now
Open source - TE
TechmemeArticle ·
Andy Jassy says AWS capacity for 2027 is "largely sold out" and a "good chunk" of 2028 capacity is also sold, though AWS reserves some capacity for on-demand (Catherine Perloff/The Information) Catherine Perloff / The Information : Andy Jassy says AWS capacity for 2027 is “largely sold out” and a “good chunk” of 2028 capacity is also sold, though AWS reserves some capacity for on-demand — Amazon Web Services capacity is mostly booked up through at least next year, Amazon CEO Andy Jassy …
Open source - SB
Shay Boloor@StockSavvyShayPost on X ·
$AMZN CEO says AI inference is still “an order of magnitude too expensive” for widespread adoption and believes Amazon’s Trainium chips could help bring costs down. And even at today’s prices AWS has already sold out its major capacity commitments for 2026 with most of 2027 and a good chunk of 2028 booked so imagine how much more demand cheaper inference could unlock.
Open source - CB
Coin Bureau@coinbureauPost on X ·
🚨JUST IN: Amazon CEO Andy Jassy reveals AWS cloud capacity is largely SOLD OUT through 2027, with a significant chunk of 2028 already booked. $AMZN is the largest cloud provider globally and has risen to its highest level this year, up nearly 30% YTD. Jassy says demand for AI computing is so strong that AWS has little capacity left to sell, strengthening the case for Amazon’s massive infrastructure spending. He warned that shortages of data centers, power and memory could persist beyond 2027, forcing Amazon to make difficult trade-offs between AI infrastructure and its core cloud business.
Open source