AI预测

• Ramp: all businesses paying for AI: 55.86% (2026-09 · 实际)
• Ramp: Information sector: 79.92% (2026-09 · 实际)
• Census BTOS: using AI now (all employer firms): 24.6% (2026-10 · 实际)
• Census BTOS: expect to use AI in 6 months: 28.6% (2026-10 · 实际)

Ramp counts businesses with a card or invoice payment for AI; the Census BTOS lines (fortnightly, all employer firms) ask whether the business used AI in the last two weeks and expects to in the next six months.

• Goldman Sachs 2026: 1.8% (预测)
• Goldman Sachs 2027: 2.5% (预测)
• Goldman Sachs 2028: 2.8% (预测)
• 收起

Goldman Sachs Research's extrapolation of its preferred estimates; it notes consensus 2027 capex projections are likely too conservative.

• Goldman Sachs 2026: $581B (预测)

Investment (capex and related spending), not only software or subscription spend.

• Gartner 2025: $982B (预测)
• Gartner 2026: $1,484B (预测)
• Gartner 2027: $1,978B (预测)
• 收起
• 此后已修订: Gartner, May 2026 (2025 $976B, 2026 $1,432B, 2027 $1,890B)

AI-optimized IaaS, servers, network fabric, AI processing semiconductors and devices; the largest segment of Gartner's AI spending forecast.

• FactSet fiscal-year floors (FY26 / FY28) 2026: $690B (预测)
• FactSet fiscal-year floors (FY26 / FY28) 2028: $900B (预测)
• 收起

Both figures are floors ("more than"), across each company's own fiscal year; FactSet's FY28 estimate is more than $900bn. Not a US-only total.

• IDC (2028) 2028: $336B (预测)

An August 2024 forecast (software, hardware and services; GenAI is $108B of it). Later Gartner and Goldman Sachs estimates are far higher.

• Gartner 2025: $1.79T (预测)
• Gartner 2026: $2.67T (预测)
• Gartner 2027: $3.64T (预测)
• 收起
• 此后已修订: Gartner, Jan 2026 (2026 $2.53T, 2027 $3.33T)
• 此后已修订: Gartner, May 2026 (2025 $1.77T, 2026 $2.6T, 2027 $3.49T)

Gartner's September 2026 forecast, its third of the year: January and May editions (grey dotted) were both revised up. The 2025 value is Gartner's estimate.

• All makers on Ramp: 81.57T (2026-10 · 实际)
• Anthropic: 48.78T (2026-10 · 实际)
• OpenAI: 30.31T (2026-10 · 实际)
• Google: 1.06T (2026-10 · 实际)

Tokens consumed per week, in trillions, by model maker. Weekly (Monday to Sunday) totals for businesses on Ramp's card and bill-pay platform, from Ramp AI Token Spend Management; a sample of Ramp customers, not the whole US. Lines are refreshed by the 6am and 6pm job.

• All makers on Ramp: $34.08M (2026-10 · 实际)
• Anthropic: $20.62M (2026-10 · 实际)
• OpenAI: $11.98M (2026-10 · 实际)
• Google: $0.64M (2026-10 · 实际)

Dollars spent on tokens per week, in millions, by model maker. Weekly (Monday to Sunday) totals for businesses on Ramp's card and bill-pay platform, from Ramp AI Token Spend Management; a sample of Ramp customers, not the whole US. Lines are refreshed by the 6am and 6pm job.

• OpenAI and Anthropic, blended: $0.57 (2026-10 · 实际)
• Anthropic: $0.64 (2026-10 · 实际)
• OpenAI: $0.48 (2026-10 · 实际)

Blended price per million tokens (token spend divided by token volume, all token types) for OpenAI and Anthropic, as Ramp charts it; weekly seven-day totals. Measured on businesses on Ramp's card and bill-pay platform, from Ramp AI Token Spend Management; a sample of Ramp customers, not the whole US. Lines are refreshed by the 6am and 6pm job.

• IDC HPC/AI Index, October 2026 2025: $28.6B (预测)
• IDC HPC/AI Index, October 2026 2030: $96.4B (预测)
• 收起

IDC October 22, 2026 HPC/AI Index: non-service-provider buyers (enterprises and governments). Publisher’s 2025 baseline and 2030 forecast, USD billions, not an observed actual series. Excludes service-provider procurement and non-AI IT infrastructure; not comparable to Gartner worldwide total AI spending or IDC US AI spending.