SemiAnalysis Cuts 2030 AI Accelerator Spending Forecast to $923 Billion
SemiAnalysis reduced its 2030 AI accelerator spending forecast to $923B, citing slower ARR growth and tighter financing. NVIDIA's computing revenue estimates were cut by 22-66% for 2027-30, while Broadcom's ASIC revenue projection fell 11% for 2026-30. U.S. data-center delays also impacted the outlook.
How this was made

The 30-second read
Why it matters
The update is a multi-year demand and pricing reset for the AI accelerator supply chain, with explicit NVIDIA-linked computing revenue cuts and a Broadcom ASIC revenue reduction driven by lower selling prices.
Market read
Traders can use the supplier-specific estimate cuts to adjust expectations for AI accelerator demand, pricing, and power-constrained deployment timelines.
What to watch
The article emphasizes ARR growth and financing availability; traders may also need to weigh potential hyperscaler reallocation, alternative funding structures, and whether power constraints shift demand toward more efficient chips rather than fewer units.
Background
SemiAnalysis revises global AI accelerator spending for 2030 downward, attributing changes to slower frontier-model lab ARR growth and tighter infrastructure financing, plus U.S. data-center capacity delays.
Ticker impact
SemiAnalysis cut NVIDIA-linked accelerator computing revenue estimates for 2027-2030, citing weaker net-new ARR growth and tighter AI infrastructure financing.
Likely pressure as investors price in reduced accelerator shipment-linked revenue and power-demand projections through 2030.
The article provides explicit, multi-year estimate cuts tied directly to NVIDIA accelerator shipments and power draw, which typically drives near-to-intermediate demand expectations.
SemiAnalysis lowered Broadcom ASIC revenue projections for 2026-30 by 11%, attributing it to lower selling prices tied to cheaper HBM.
Likely modest downside bias as lower ASP assumptions offset stable unit volume in the forecast.
The article quantifies a multi-year revenue projection cut and links it to pricing, but does not provide a direct catalyst like guidance or contract changes.
Market effects
Signals a demand and financing headwind for AI accelerator ecosystems, with particular sensitivity for projects lacking hyperscaler backing.
U.S. data-center capacity delays (5.8 GW) reinforce near-to-intermediate timing risk for AI infrastructure buildouts.
Could pressure global AI capex expectations and supplier revenue models across the accelerator and networking stack through 2030.
Counterpoint
Forecast cuts may reflect financing timing and capacity phasing rather than a structural demand collapse, so supplier earnings could be less impaired than the long-range model suggests.
Key entities
- research_firmSemiAnalysis
Analyst firm issuing the revised AI accelerator spending outlook and supplier estimate cuts.
- supplierNVIDIA
Accelerator supplier whose shipment-linked computing revenue estimates were cut sharply for 2027-2030.
- supplierBroadcom
ASIC supplier whose 2026-30 revenue projection was lowered, with shipment volume unchanged.
- supplierAmazon
Trainium accelerator outlook for 2027-2028 held steady in the revision.



