# data center revenue — X 热门讨论 (2026-09-22 00:02 UTC)
## @trevornoren (Trevor Noren) · 09-21 17:11 · ♥40 ↻17 💬3 JPM: "Anthropic began the year with 1-2 GW of compute but according to a Sept 6 report has now committed to pay for access to another 15 GW of power that it will access over the next few years. The estimated cost: ~$500b to rent cloud capacity, chips and leases on data centers. This compares to $180b of projected data center expense that Anthropic mentioned last year through 2029. Anthropic is catching up to OpenAI which increased its estimate of data center expense from $660b to $750b through 2030, which is shooting for 30 GW by 2030 (equivalent to 30 modern 1 GW nuclear reactors, and the US only has ~95 of them)."
Of course, the first concern in this is whether Anthropic and OpenAI can bring in the average revenue per customer needed to offset spending commitments. As I warned in my recent report on "The AI Trade" (https://t.co/wQQNniS2kj): "We often pose the same question to clients: What if agentic AI is more akin to private flight than cell phones? The TAM for both private flight and cell phones is the entire global population. However, scale made cell phones affordable for most of the population while that remains fundamentally impossible for private flight. What if agentic only makes cost sense for a sliver of use cases across sectors? We believe it’s a question far too few are confronting today."
The second concern is whether these commitments actually yield the compute capacity needed. To again quote the report: "Our more immediate concern is not data center capacity outpacing compute demand. Instead, it’s hyperscalers failing to build out the data center capacity required to improve their models enough to inspire adequate enterprise spending.
The data center buildout is falling behind schedule. This year, between 30 and 50% of the large-scale data center capacity that was expected to come online will likely be delayed. As for 2027, more than 60% of data-center capacity planned for completion wasn’t yet under construction as of May, and another 7% had been officially delayed. Delays have always been part of the data center construction reality—historically, only 72% of data center capacity has come online on time. However, things are clearly getting worse. Goldman expects only about half of the AI computing capacity scheduled to activate between now and 2028 via data center construction will actually come online by its target date. As asset valuation firm Barkr calculated in an August report, delays at that rate would translate to a compute supply/demand gap of 13.4 to 19.2 GW in 2027 and 27.3 to 36.8 GW in 2028."
As for electricity supply: "Annual growth between 2% and 3% is expected through the end of the decade, with data center electricity demand more than doubling from 31 GW in 2025 to 66 GW in 2027. BofA has projected that the US will face an electricity generation shortfall of 100 GWs or more by 2030, with only roughly 90 GWs of new supply coming against more than 200 GWs of new demand."
JPM link: https://t.co/Hk8xGFZKN2 https://x.com/trevornoren/status/2102083375494492579
## @thatsKAIZEN (Kaizen D. Asiedu) · 09-21 20:39 · ♥35 ↻2 💬5 Clear Thoughts: The Press Ban, Trump's $5K Promise, Macklemore “Genocide”, and the Lone Clancy Juror https://x.com/thatsKAIZEN/status/2102135704235216904
## @Venu_7_ (Venu) · 09-21 23:15 · ♥36 ↻2 💬3 $HPE with a beautiful cup + handle here.
AI infrastructure + networking play, with Cloud & AI revenue +25% and Data Center Networking +112% last quarter.
Now knocking on the $64 breakout level. https://t.co/hDMn4cz0ss https://x.com/Venu_7_/status/2102174881429569975