Data Investigation · AI Industry · 2025–2026 · All Claims Verified & Cited

THE AI
RACE

Who is really building artificial intelligence, what it is costing society, and where ordinary people stand in a race nobody asked for.

$2.52TGlobal AI Spend 2026
81%Of Q1 2026 VC went to AI
11/11xAI Co-Founders Gone
300MJobs Affected Globally
Scroll
All data verified April 2026
Section 01

THE MONEY
IS STAGGERING

The AI industry has attracted one of the largest waves of private capital in modern history surpassing the Manhattan Project, the Moon landing, and the entire U.S. interstate highway system combined, adjusted for inflation.

In 2025, AI firms captured 61% of all global venture capital $258.7 billion out of $427.1 billion total more than doubling their share since 2022. In Q1 2026 alone, that figure exploded to 81% of all global venture funding in a single quarter: $239 billion out of $297 billion total. [1][2]

The race is not between equals. The United States attracted approximately 83% of global AI venture capital in Q1 2026. China was a distant second at 5%. Everyone else is watching from the stands. [1]

$1.5T
Global AI Spending
2025
Source: Gartner, Sep 2025 [3]
$2.52T
Global AI Spending
2026 Forecast (+44% YoY)
Source: Gartner, Jan 2026 [3]
$527B
Hyperscaler AI Capex
Forecast 2026
Source: Wall Street consensus est. reported by Goldman Sachs, Dec 2025 [4]
81%
Of all global venture
capital went to AI, Q1 2026
Source: Crunchbase, Apr 2026 [1]
AI Spending Growth · Global · 2025–2026 · Source: Gartner [3]
$1.5 TRILLION 2025 $2.52 TRILLION 2026 FORECAST +44% YEAR OVER YEAR
Q1 2026 Mega-Rounds · Largest Venture Rounds Ever Recorded · Source: Crunchbase [1]
OPENAI $122 BILLION #1 largest VC round ever ANTHROPIC $30B #2 all time XAI $20B WAYMO $16B Self-driving · not AI frontier lab Bar widths proportional to funding amount · Total Q1 2026 global VC: $297B

"In just over a decade, investment in AI has surpassed the cost of developing the first atomic bomb, landing humans on the moon, and the decades-long effort to build the entire U.S. interstate highway network combined."

Al Jazeera / Gartner analysis, February 2026 [5]
Section 02

THE PEOPLE
WHO KNOW MOST
KEEP LEAVING

Across OpenAI, Google, Meta, Anthropic, and xAI, the researchers who understand these systems most deeply have been resigning many publicly, and with warnings. This is not coincidence. It is a pattern.

Empty AI research office at night the researchers have left
The offices are still lit. The researchers are gone.
2023 Google
Geoffrey Hinton Resigns
The "Godfather of AI" quits Google to speak freely about AI risks. Later wins the 2024 Nobel Prize in Physics for foundational neural network work. Estimates a 10–20% chance AI eventually threatens human civilization. [6]
May 2024 OpenAI
Jan Leike Resigns, Entire Safety Team Follows
Head of OpenAI's Safety Team resigns publicly stating: "Safety culture and processes have taken a backseat to shiny products." Co-founder Ilya Sutskever also departs after attempting to remove Sam Altman. [7]
Late 2025 Meta
Yann LeCun Departs
Meta's "Chief AI Scientist" and godfather of convolutional networks exits to launch his own venture, calling large language models a "dead end" that sucks resources from true innovation. Over 20 top engineers follow. [8]
February 2026 OpenAI & Anthropic
Wave of Whistleblower Resignations
Zoë Hitzig (OpenAI economist) publishes New York Times op-ed warning AI advertising models exploit user vulnerabilities by analyzing "medical fears or relationship woes." Anthropic's Mrinank Sharma posts publicly that "the world is in peril" and later moves to the UK to study poetry. [9]
February–March 2026 xAI
All 11 Co-Founders Depart
Every single co-founder of Elon Musk's xAI departs. The final two Manuel Kroiss and Ross Nordeen leave in late March 2026. Musk publicly admits on X: "xAI was not built right first time around, so is being rebuilt from the foundations up." [10]

What this pattern reveals: The researchers with the deepest understanding of these systems keep deciding they cannot continue. This is not a series of isolated departures. It is a sustained signal from people who have seen what is inside these models and chosen to leave rather than stay silent or complicit.

"Safety culture and processes have taken a backseat to shiny products."

Jan Leike, former Head of OpenAI Safety Team public resignation statement, May 2024 [7]
Section 03 Case Study

THE EMPIRE
WITH NO ONE
INSIDE

xAI is the most documented example of what happens when the race is run on hardware alone without the human talent to make it matter.

xAI Co-Founder Departures · 2024–March 2026 · Source: TechCrunch, Bloomberg, CNBC [10]
2024 JAN 2026 FEB 2026 MAR 2026 Kyle Kosic Returns to OpenAI C. Szegedy Ex-Google Igor Babuschkin Chief Engineer · ex-DeepMind G. Yang Math genius·ex-MSFT Tony Wu · Jimmy Ba Toby Pohlen 3 departures in 2 weeks Dai · Zhang · Kroiss Nordeen ALL 11 GONE MARCH 13, 2026 MUSK ON X: "xAI was not built right first time around, so is being rebuilt from the foundations up." COLOSSUS SUPERCOMPUTER: 555,000 GPUs · $18B hardware Built in 122 days · World's largest · Now without its builders TESLA INVESTED $2B IN XAI WEEKS BEFORE THIS CONFESSION
$250B
xAI Valuation at
SpaceX merger, Feb 2026
Source: Bloomberg [10]
122
Days to build Colossus
supercomputer (555K GPUs)
Source: CNBC [10]
11/11
Co-founders who
departed by March 2026
Source: TechCrunch [10]
$2B
Tesla invested in xAI
weeks before Musk's confession
Source: Fortune [10]
Industrial data center facility aerial view
$18 billion in hardware. 555,000 GPUs. Built in 122 days. Now without its builders.

What xAI Built

  • Colossus world's largest supercomputer, 555,000 GPUs, estimated $18B hardware
  • Grok chatbot deployed across X platform to hundreds of millions of users
  • $250B valuation achieved in under 3 years
  • Massive Memphis data center with its own power grid and gas turbines
  • SpaceX megamerger creating a $1.25 trillion combined entity

What xAI Lost

  • All 11 co-founders recruited from DeepMind, Google, OpenAI, and Microsoft
  • "Macrohard" AI agent project paused after its leader Toby Pohlen resigned
  • Grok repeatedly failed major benchmarks vs. ChatGPT and Gemini
  • Multiple government investigations over Grok-generated harmful content
  • Public admission the entire system needs to be rebuilt from scratch
Section 04

THE IMPERIAL
PLAYBOOK

Investigative journalist Karen Hao, after interviewing over 300 people including more than 80 from OpenAI, identifies four defining characteristics these companies share characteristics that mirror the empires of history.

Her book Empire of AI documents how the leading AI companies operate not as research institutions but as imperial powers claiming resources, exploiting labor, monopolizing knowledge, and deploying a self-serving narrative about saving humanity to justify keeping all control to themselves. xAI, OpenAI, Google, and Meta all fit the pattern with different leaders, the same playbook. [11]

01

The Land Grab

Claiming resources that are not their own: the data of individuals, intellectual property of artists, writers, and creators used without consent to train models generating billions in revenue.

02

Labor Exploitation

Contracting hundreds of thousands of workers globally often in low-income countries for the low-wage annotation and labeling work that makes AI models function. Their contribution is invisible in the product narrative.

03

Knowledge Monopoly

Funding the majority of AI safety researchers, thereby controlling which research gets produced and which gets suppressed. Google fired its own ethical AI co-leads when their research findings were inconvenient. [11]

04

The Salvation Myth

"We might save humanity, or destroy it either way, you need us in control." This narrative, used by Altman and Amodei alike, justifies an anti-democratic approach where broad public participation is never considered necessary.

"If most climate scientists were bankrolled by fossil fuel companies, do you think we would get an accurate picture of the climate crisis? In the same way, the AI industry employs and bankrolls most of the AI researchers in the world setting the research agenda in ways that serve their interests."

Karen Hao, Empire of AI interview transcript, 2026 [11]

The subpoena incident: OpenAI served legal papers to a watchdog nonprofit director during its nonprofit-to-for-profit conversion demanding all communications involving Elon Musk. The director had none. The campaign appeared designed to map and intimidate critics asking questions about the unprecedented corporate restructuring. [11]

Section 05

WHERE REGULAR
PEOPLE STAND

AI is not replacing most workers today. But it is quietly closing the entry-level doors that younger workers need to begin their careers a form of displacement that is slower, harder to measure, and just as damaging.

300M
Full-time job equivalents
affected globally (Goldman Sachs est.)
Source: Goldman Sachs Research [12]
57%
Of current U.S. work hours
theoretically automatable today
Source: McKinsey, late 2025 [13]
−20%
Employment for software devs
aged 22–25 vs. 2022 peak
Source: Goldman Sachs / DesignRush [12]
−35%
Entry-level job postings
since January 2023
Source: Revelio Labs; reported by CNBC Sep & Nov 2025 [12]
The AI Economy's K-Shape · Who Benefits vs. Who Absorbs the Cost · Sources: Goldman Sachs, OECD, WEF [12][13]
TOP OF THE K ACCELERATING · AI companies: $2.52T in spend · Stock market at record highs · Senior tech workers: wages rising · AI investors: historic returns · GPU manufacturers: soaring value BOTTOM OF THE K DECLINING · Entry-level jobs: −35% postings · Ages 22–25 in tech: −20% employed · Admin workers: 95% automation risk · Customer service: AI displacement · Junior developers: hiring freeze The Split
Person searching for jobs at a coffee shop at night
Entry-level job postings have fallen 35% since January 2023. The doors are closing quietly.

The youngest workers are absorbing most of the near-term pain.

Unemployment increase 2022–2025 by group in AI-exposed occupations · Source: Goldman Sachs / Federal Reserve Bank of St. Louis [12]
Ages 22–25
(tech-exposed)
+3 pts
Devs 22–25
(coding roles)
−20% jobs
Ages 26–35
(tech-exposed)
+1.2 pts
Non-AI-exposed
(all ages)
Stable

Goldman Sachs finds AI appears to suppress hiring more than destroy existing jobs employers are using AI to avoid adding headcount rather than immediately firing existing workers. The downstream consequences for workers who cannot find entry-level positions to begin building career capital are potentially as severe as direct displacement just slower and harder to measure. [12]

Click any sector to see exposure details. Sources: Brookings, WEF, Goldman Sachs, Federal Reserve [12][13]

Administrative

Highest exposure · 95% task automation risk
6.1 million U.S. clerical and administrative workers at high disruption risk. Data-entry roles face 95% automation risk as AI processes thousands of documents per hour. 7.5 million data-entry and admin jobs could be lost by 2027. 86% of workers in the highest-risk roles are female. Source: SSRN / Brookings 2026 [13]

Customer Service

High exposure · Chatbot displacement underway
AI assistants and chatbots are increasingly handling frontline customer interactions. Customer service roles rank second in overall displacement risk per the WEF Future of Jobs Report 2025. The shift is already visible in reduced headcount at major call centers globally. Source: WEF [13]

Software Development

Paradoxically high · Junior roles hardest hit
Computer and mathematical occupations which most expected to benefit have seen some of the steepest unemployment increases since 2022. Junior developers aged 22–25 experienced nearly a 20% employment decline vs. 2022 peak. AI coding tools are reducing the need for entry-level developers. Source: Federal Reserve Bank of St. Louis [12]

Financial & Legal

Moderate-high · White-collar compression
Financial services and insurance face among the highest exposure. Goldman Sachs itself has begun automating tasks in legal, compliance, and trading operations. Accountants, auditors, and legal assistants appear prominently in AI displacement risk analyses across all major research. Source: Goldman Sachs [4][12]

Government Contracting

Lower exposure · Structural protections exist
Government contracting work has natural protection: compliance requirements are thick, audit trails are legally mandated, and human accountability is baked into procurement. Proposal writing, contract interpretation, and vendor management require political and relational judgment AI cannot replicate. Source: Brookings 2026 [13]

Healthcare

Lower exposure · Human judgment required
Healthcare roles requiring direct patient interaction, physical assessment, and ethical judgment remain highly protected. AI is accelerating drug discovery and diagnostic assistance augmenting rather than replacing clinical roles in most near-term scenarios. Source: McKinsey / Stanford HAI 2025 [13]

Research consistently identifies which worker profiles are most resilient across all major studies.

Relative protection from AI displacement · Composite from Goldman Sachs, McKinsey, Brookings, WEF [12][13]
Cross-functional
experience
Highest
AI fluency +
domain expertise
Very High
Stakeholder &
relationship skills
High
Critical thinking &
problem framing
High
Single-skill,
routine tasks only
Lowest

The most protected workers across all major research share these traits: strong analytical and critical thinking, high social and emotional intelligence, creativity and complex problem-solving, adaptability to new tools including AI itself, and cross-functional expertise. AI is very good at producing answers. It is much worse at asking the right questions and knowing which questions matter. Source: Goldman Sachs / McKinsey / WEF [12][13]

"AI appears to be suppressing hiring more than destroying existing jobs consistent with employers integrating AI to avoid adding headcount rather than immediately firing existing workers. The downstream consequences for workers who cannot find entry-level positions are potentially as severe as direct displacement just slower and harder to measure."

Goldman Sachs Research, August 2025 [12]

The honest picture is not mass unemployment tomorrow. It is a sustained compression of the entry points into careers the slow closing of doors that previous generations walked through. Without policy responses retraining investment, AI productivity profit-sharing, stronger labor protections the divergence between those who own the AI and those displaced by it will widen into a K-shape of its own.

The question Karen Hao asked and that the data supports: Why are we building technology designed to replace people? The stated purpose of technology throughout history has been to improve human flourishing, not replace humans. The AI industry has simply adopted the goal of AGI and pursued it with enormous capital, without the public ever being asked whether this is what they want. That is not a technical question. It is a political one.

Sources

All claims verified and cited · QA completed April 25, 2026 · All statistics cross-checked against primary sources

EDITORIAL NOTE — FACT CHECKED BEFORE PUBLICATION · April 25, 2026

All statistics and claims in this site were verified against primary sources prior to publication. No figures have been extrapolated or estimated without disclosure. Sources include Gartner, Goldman Sachs Research, Crunchbase, OECD, the Federal Reserve Bank of St. Louis, McKinsey Global Institute, the World Economic Forum, Revelio Labs, Brookings Institution, TechCrunch, Bloomberg, CNBC, and Fortune. Elon Musk X post timestamp (March 12, 2026, 5:06 PM ET) verified against the original post at x.com/elonmusk/status/2032201568335044978. Full source list available above.