Research Analysis / March 2026 / 33 data points

The AGI Incentive Map

Who predicts AGI is imminent, who doesn't, and what financial incentives lie behind each position. The correlation is real — but the full picture is more nuanced than "follow the money."

The dividing line isn't money vs. no money. It's whether your specific valuation collapses if AGI is 20 years away. Model builders say 2027. Platform players with equal exposure say "quite far." Researchers say 2040+.
$1.34T
Private valuations of companies claiming "AGI by 2027" (OpenAI + Anthropic + xAI)
20 yrs
Gap between model-builder predictions (~2027) and researcher survey median (~2047)
76%
AI researchers who say scaling current approaches to AGI is "unlikely" (AAAI, n=475)
5 of 5
Platform CEOs (Nadella, Pichai, Jassy, Huang, Gates) who are more cautious than the labs they fund

The Correlation (n=33)

Each dot is a person or group. X-axis: financial exposure to AGI arriving soon. Y-axis: how soon they predict it. The four quadrants tell four different stories. Hover for details.

Model builders
AGI investors
Platform pragmatists
Safety believers
Independent researchers
Aggregate data
MODEL BUILDERS & BACKERS
RESEARCHERS & ACADEMICS
SAFETY BELIEVERS
PLATFORM PRAGMATISTS

Three Positions, One Technology

The conversation splits into three distinct camps — each with a different incentive structure.

Selling AGI
"We are now confident we know how to build AGI."
Sam Altman, Jan 2025 — $730B
"A country of geniuses in a datacenter, as early as 2026."
Dario Amodei — $380B
"AGI by 2027 is strikingly plausible."
Leopold Aschenbrenner — $5.5B fund
"I was born to realise ASI."
Masayoshi Son — $41B in OpenAI
Investing, Not Hyping
"AGI milestones are nonsensical benchmark hacking."
Satya Nadella — 27% of OpenAI
"We are quite far from a generalized technology."
Sundar Pichai — CEO, $4T Alphabet
"It's so early right now in the AI space."
Andy Jassy — $50B in OpenAI
"Whether we reach that point in a decade or a century."
Bill Gates — Microsoft advisor
Studying AI
"People bloviating about AGI in a year or two. Completely delusional."
Yann LeCun, Turing Award — left Meta
"AI agents are slop. They just don't work."
Andrej Karpathy — ex-OpenAI
"Decades away. AGI hype misleads students and CEOs."
Andrew Ng — DeepLearning.AI
"Revised from 2027 to 2030 when evidence warranted."
Daniel Kokotajlo — forfeited $1.7M
The Zuckerberg-LeCun test

The most instructive case: Zuckerberg ("Superintelligence is now in sight") and his own chief scientist LeCun ("Completely delusional") work at the same company. The money-holder and the scientist have opposite positions. This pattern repeats: Altman vs. Karpathy, Son vs. Nadella, Hassabis vs. Pichai. In every case, the person controlling capital is more bullish than the person doing the research.

The Money Behind the Claims

Three model-building companies worth $1.34T. Their platform partners — with equal or greater exposure — won't endorse their timelines.

OpenAI
$730B
Revenue: $13B · Loss: $14B · 56x
Amazon's $35B tranche requires "achieving AGI or IPO." Projected $665B cumulative burn through 2030.
Says AGI: 2027
Microsoft 27% SoftBank $41B Amazon $50B Nvidia $30B
Anthropic
$380B
Revenue: $14B ARR · 27x
$67.3B raised across 17 rounds. Preparing IPO 2026. The only company with an official AGI date.
Says AGI: 2026-2027
Amazon $8B+ Google 14% Nvidia $10B Goldman Sachs
Microsoft (partner)
27% of OpenAI
~$13B invested · $80B AI capex
CEO Nadella calls AGI milestones "nonsensical benchmark hacking" and warns of AI infrastructure "overbuild."
Says AGI: won't commit
Also invested $5B in Anthropic
The structural difference

Model builders (OpenAI, Anthropic, xAI, SSI) have valuations that collapse if AGI is 20 years away. Platform players (Microsoft, Google, Amazon, Nvidia) profit from AI adoption at any pace. This explains why two people can have equal financial exposure to AI but radically different AGI claims. Nadella's $13B in OpenAI doesn't need AGI to be imminent — Azure grows either way. Altman's $730B valuation does.

The Signal (Controlled for Incentives)

What does the evidence look like when you separate the signals by incentive structure?

Independent researchers & aggregate data (no AGI-dependent valuations)
  • LLMs are not a path to AGI. Stated by Turing Award winners (Sutton, LeCun), an OpenAI co-founder (Karpathy), and top academics. "A dead end" for general intelligence. — Sutton, LeCun, 2025
  • 76% of AI researchers say scaling current approaches to AGI is "unlikely." 77% prefer practical AI over direct AGI pursuit. — AAAI, 475 researchers
  • Median researcher prediction: 2047. Down from 2059 in 2022, still 20 years later than CEO claims. — AI Impacts, 2,778 researchers
  • $700B/yr AI spend = "basically zero" GDP growth. 95% of businesses trying AI found zero value. — Goldman Sachs, MIT Tech Review
  • Prediction markets shifted later in 2025-2026. Both 25% and 50% dates pushed back ~2 years. — Metaculus, Feb 2026
The complication: safety believers (low financial stake, believe it's near)
  • Geoffrey Hinton (Nobel, left Google): "5-20 years." Progress "even faster than I thought." Modest AI investments but reputation-staking alarm. — 2025
  • Yoshua Bengio (Turing Award): 2028-2043 at 90% confidence. Calls for suspension of AGI development. Academic salary, no lab equity. — 2025
  • Daniel Kokotajlo (ex-OpenAI): Forfeited $1.7M to speak freely. Revised from 2027 to 2030 when evidence warranted. The strongest "honest calibration" data point. — 2025-2026
  • These voices prevent a simple "only money explains it" story. But note: their timelines (2028-2043) are far wider than the model builders' (2026-2027), and several have career/institutional incentives tied to AI being dangerous.

Timeline comparison by incentive category

Model builders & direct backers (Altman, Amodei, Musk, Son)
~2027
Safety believers, low financial stake (Hinton, Bengio, Kokotajlo)
~2030
Prediction markets (Metaculus community)
~2033
Platform pragmatists (Nadella, Pichai, Jassy, Gates)
~2035+
Independent researchers (LeCun, Karpathy, Ng, Brooks)
~2037
AI researcher survey median (n=2,778)
~2047

Bar length = years from today. Sources: AI Impacts, Metaculus, AAAI, public statements.

The Bottom Line

The technology is real and useful. LLMs are generating genuine revenue. Nobody credible disputes this.

The "money = hype" correlation is real, with caveats. Model builders (OpenAI, Anthropic, xAI) whose valuations depend on AGI proximity cluster at 2026-2027. Platform players (Microsoft, Google, Amazon) with equal or greater AI exposure but no AGI-dependent valuations cluster at "too early to say" or "quite far." This isn't coincidence.

But financial incentive isn't the only explanation. A cluster of safety researchers — Hinton, Bengio, Kokotajlo — believe AGI is near-ish (2028-2035) without having billions at stake. Their timelines are wider and more hedged than the model builders', and some have revised longer when evidence warranted, but they exist. The honest framing: financial incentives explain the extremity and precision of the claims (exactly 2027, total confidence), not necessarily the direction.

The most telling signal: In every case where a capital-controller and a researcher exist at the same organization, the capital-controller predicts AGI sooner. Zuckerberg vs. LeCun. Altman vs. Karpathy. Son vs. Nadella. Hassabis vs. Pichai. This pattern has zero exceptions in our dataset.

The controlled estimate: Strip out everyone whose valuation depends on AGI proximity. The remaining voices — platform CEOs, academics, Turing Award winners, 2,778 surveyed researchers, prediction markets — converge on a range of 2030-2047, with most weight in the 2033-2040 band. The technology will continue to improve. AGI, however defined, is probably a decade-plus away, not one to three years.

Predicts AGI
Financial exposure