Browse
Investor Perspectives Tracker
What the money is saying about AI markets: theses from VC firms, bank research, and independent analysts — one row per perspective, with the key numbers, the stance, and a link to the original post or report. The same row-per-item format as the Policy Tracker. Updated as new letters, reports, and notes land.
Share
Vintage: 2026-10. Added October 1, 2026: David George (a16z) on OpenAI's distribution moat — models win because of distribution, not the other way around (Bullish, Enterprise adoption); Morgan Stanley's AI capex deceleration note — 93%→61%→14% capex growth into 2028, $286B in delayed projects, 57 GW power shortfall (Cautious, Capex & earnings). Launched September 30, 2026 with 12 perspectives: a16z's State of Markets II (the bits-to-atoms rotation, GPU pricing climbing instead of decaying, ~2% of the S&P 500 tracking AI metrics), Sequoia's AGI and 99.9%-cognitive-work theses, Chamath Palihapitiya's land-power-shell bull case and his 'tokenmaxxing' warning, Mary Meeker's Trends: AI, Goldman Sachs / JPMorgan / Morgan Stanley research on the $7.6T capex arc and the 38 GW power gap, and McKinsey's 11M-worker displacement read.
14 of 14 rows
| Entity | Metric | Value | Period | Confidence | Source |
|---|---|---|---|---|---|
Andreessen Horowitz (a16z) Capex & earnings | State of Markets II — David GeorgeNew Tech contributed ~76% of S&P 500 earnings growth in 2026 (late Aug) — 'tech is the everything cycle now.' Rotation from bits to atoms: the AI buildout is surging demand for semiconductors, power, and networking; hyperscaler free cashflow is being converted into semiconductor free cashflow, increasingly debt-funded. Compute demand still outpacing supply — even A100 GPUs pricing at or above January levels. Adoption broad but shallow: ~30% of the S&P 500 report 'quantifiable impact' of AI, only ~2% report a tracked metric; ~2% of US households paying for AI (April). | Bullish | Sep 2026 | Press-reported | [1] |
Andreessen Horowitz (a16z) Valuations | State of Markets — first editionNew ~55% of AI growth concentrated in private markets; AI companies grew at a median 106% YoY in 2025 vs 39% for non-AI; top decile hit 693% YoY. AI stocks accounted for ~78% of S&P 500 returns since Nov 2022. Hyperscalers allocating 30–65% of revenue to capex — spending could reach 75% of operating cash flow in 2026. | Bullish | Jan 2026 | Press-reported | [2] |
Sequoia Capital Enterprise adoption | '2026: This is AGI' — Pat Grady & Sonya HuangNew AGI 'functionally here' via long-horizon agents — three compounding ingredients: pre-training (2022/ChatGPT), inference-time compute (o1, late 2024), long-horizon agents (2026). METR data: AI's ability to complete long-horizon tasks doubling every ~7 months — a day's expert work by 2028, a year's by 2034. AI apps of 2026–27 will be 'doers, not talkers': users go from working as ICs to managing teams of agents. | Bullish | Jan 2026 | Press-reported | [3] |
Sequoia Capital Enterprise adoption | The cognitive-revolution thesis — Konstantine BuhlerNew Machines will do 99.9% of the world's cognitive work — the same share of physical work captured between 1800 and today, only faster. The prize: ~$10T in services revenue that software never reached, vs a ~$650B software market ('sell outcomes, not tools'). PitchBook: AI startups raised $407B in H1 2026 (vs $264B all of 2025); AI absorbed 86% of the $412.7B deployed. Warns of an 'Engels' pause' compressed into five years: after 1780, output per worker rose 46% while real wages rose just 12%. | Bullish | Sep 2026 | Press-reported | [4] |
Social Capital (Chamath Palihapitiya) Energy & grids | AI investing guide — 'land, power, shell'New LPS (land, power, shell) is 'still the most obvious and fastest path to cash on cash returns' — 'energized land can explode in value.' He and partner Anita Verma-Lallian have locked in close to 6 GW of power through 2029. Durable margins belong to 'harnesses' and the applications built on top; zoning-approved land and silicon access give owners negotiating leverage over everyone downstream. | Bullish (infrastructure) | Aug 2026 | Press-reported | [5] |
Chamath Palihapitiya Enterprise adoption | The 'tokenmaxxing' earnings warningNew Unchecked AI-token spending will cause surprise EPS misses of 'a few pennies' — 'CEOs and CFOs probably have no idea how much tokenmaxxing is going on inside their organizations.' His own startup (8090) was trending toward >$10M/year in AI usage credits (Mar 2026). Nobody has shown 'repeatable, audited, verifiable ROI' even as capex and opex climb with token costs. | Cautious | Jul 2026 | Press-reported | [6] |
BOND Capital (Mary Meeker) Chips & compute | Trends — Artificial IntelligenceNew 340 slides; AI moving faster than any prior technology transition. ChatGPT hit 800M users in 17 months (1M in 5 days vs the iPhone's 74). Big Six capex rose 63% YoY to $212B in 2024; NVIDIA GPUs improved 225x in performance over 8 years. Training datasets growing 260%/year, training compute 360%/year; AI job postings up 448% since 2018. Meeker: 'a period for lots of wealth creation and wealth destruction.' | Bullish | May 2025 | Press-reported | [7] |
Goldman Sachs Global Investment Research Capex & earnings | The AI revenue-hurdle mathNew Six US hyperscalers need ~$1.42T in cumulative 2028–2030 revenue (~$11.6B of annual revenue per GW of compute, at 15% ROIC) to justify 2026–27 'Phase II' AI compute capex — ~$300B/year just to break even, ~$1T/year for 'satisfying' returns. The offset case: $1.69T in cloud backlog. Cumulative AI infra capex of ~$7.6T over 2026–31. | Mixed | Sep 2026 | Press-reported | [8] |
J.P. Morgan Capex & earnings | Financing AI infrastructure and U.S. data centersNew Five largest US hyperscalers' capex to reach $697B in 2026 — up $173B from start-of-year projections. >$5T to be invested globally between 2026–2030 in data centers, AI infrastructure, and related power networks; power supply is 'the most critical factor.' JPM originated $9.6B across two construction loans for Stargate's Abilene campus (Stargate: up to $500B over four years). 'AI financing is the biggest secular theme in our professional lifetimes.' | Bullish | Sep 2026 | Press-reported | [9] |
Morgan Stanley Research Capex & earnings | AI returns and the 2028 capex slowdownNew Model companies using their own compute for API services can reach 46% ROIC; ~25% even on third-party infra — 'AI investment is gradually developing the ability to fund itself.' Combined data-center capex: $917B (2026) → $1.47T (2027, ~60% YoY) → growth slows to ~12% in 2028; capital rotates from hardware/semiconductors toward enablement and software — the mobile-era pattern. | Bullish | Sep 2026 | Press-reported | [10] |
Morgan Stanley Research Energy & grids | The 38 GW power shortfallNew US data centers need ~68 GW of new power capacity between 2026–2028; ~30 GW is covered by construction and contracted grid capacity — the remaining ~38 GW 'has no clear answer yet.' Grid interconnection requests take 5–7 years in some regions vs hyperscalers' 2–3-year timelines. Behind-the-meter gas turbines could add 15–20 GW. | Neutral | Aug 2026 | Press-reported | [11] |
McKinsey Labor markets | AI workforce displacement readNew ~11M US workers may need occupation switches by 2035; 36M jobs cut vs 40M created — net positive, but the churn is the story. Low-wage workers 8x more likely to need to switch occupations. | Neutral | Sep 2026 | Press-reported | [12] |
Andreessen Horowitz (a16z) — David George Enterprise adoption | OpenAI Understands Something Important and RareNew The purpose of a company is to create customers: OpenAI will keep the best models, chips, and products because it has the best distribution, not the other way around. New behavior and distribution beat pricing and switching costs as durable moats. Developer routing share OpenAI vs Anthropic went ~20% → 50% on OpenRouter since June 2026; OpenAI's own 'signals' (Aug 12): 95%/93% plugins/skills vs 21%/19% frontier vs 9%/3% typical — the installed base is already living inside OpenAI's platform. | Bullish | Sep 2026 | Press-reported | [13] |
Morgan Stanley Research Capex & earnings | AI capex growth slowdown (Sep 2026 note)New The AI investment cycle shifts from hardware-led to software/applications by 2028: hyperscaler/neocloud capex YoY growth falls 93% (2026) → 61% (2027) → 14% (2028). The physical drag: ~$286B of US data-center projects delayed or cancelled since 2025 (local opposition) and a 57 GW power shortfall 2026–28 — a 30–40% shortfall. Compute demand still exceeds supply for years; monetization becomes the test. AI contributed 0.6pp to US GDP in 2025, 0.8pp in 1H26. | Cautious | Sep 2026 | Press-reported | [14] |
Sources
- State of Markets II — a16z (David George) (a16z)
- a16z State of Markets report: 55% of AI growth in private markets (MPost)
- 2026: This is AGI — Sequoia (Pat Grady & Sonya Huang) (Sequoia Capital)
- Sequoia's Konstantine Buhler bets AI will claim 99.9% of human thinking (Startup Fortune)
- Palihapitiya harnesses apps: AI investing guide (land, power, shell) (Cryptopolitan)
- Chamath Palihapitiya: AI 'tokenmaxxing' earnings warning (Crypto Briefing)
- Trends — Artificial Intelligence (May '25) — BOND (Mary Meeker) (BOND Capital)
- Goldman Sachs calculates the revenue hyperscalers need to justify AI capex (Ranzware)
- Financing AI infrastructure and U.S. data centers — J.P. Morgan (J.P. Morgan)
- Morgan Stanley: AI returns and the 2028 capex slowdown (BigGo Finance)
- Morgan Stanley: AI data centers face 38 GW power gap (Crypto Briefing)
- McKinsey says AI could force 11 million US workers to switch occupations (Seoul Economic Daily)
- OpenAI Understands Something Important and Rare — a16z (David George) (a16z)
- AI spending growth seen slowing sharply by 2028 — Morgan Stanley capex deceleration note (The Hindu BusinessLine (ANI, citing Morgan Stanley))
How to cite this page
SundayPyjamas Impact Foundation. "Investor Perspectives Tracker." Impact Suite. Accessed October 1, 2026. https://sundaypyjamas.org/tracker/investor-perspectives.