Related Insight
AI Stocks: Who Actually Won Since ChatGPT?
NVDA +800% (2022 trough to 2024 peak). SK Hynix +2,000%. Who won the rally — and what the four waves mean for what comes next.
→
Insights — Q2 2026
The AI Bill:
What the World Has Gained and Lost Since 2022
Since ChatGPT launched in November 2022, the world has poured over $1.2 trillion into AI infrastructure, consumed more electricity than Saudi Arabia, and placed 85 million jobs at risk. The gains are real — and so are the costs. Here's the most complete data scorecard of what AI has actually delivered, what it's taken, and where the ledger stands today.
$1.2T+
Big Tech AI capex since ChatGPT (2022–2025)
Epoch AI / Visual Capitalist, 2026
448 TWh
Electricity consumed by global data centers in 2025
UNU-INWEH / IEA, June 2026
~Zero
AI's measured contribution to official US GDP in 2025
Goldman Sachs Chief Economist, Feb 2026
85M
Jobs projected at risk globally from AI by end of 2026
World Economic Forum, 2025
The Full Ledger
The Scorecard: AI's Gains and Costs at a Glance
Three and a half years in, AI's ledger is deeply asymmetric. The costs are concentrated, measured, and present. The gains are diffuse, partially unmeasured, and still accruing. This is not a reason for pessimism — it is the pattern of every major general-purpose technology. But it is a reason to be honest about where we are in the cycle.
| Category | Gain / Benefit | Cost / Risk | Verdict |
|---|---|---|---|
Capital Investment 2022–2025 |
$1.2T+ deployed; largest infra cycle since telecoms | ROI not yet visible in GDP; ~50% imports dilute domestic gain | ⚖ Pending |
Productivity Measured gains |
10% productivity lift in AI-exposed industries; 56% wage premium for AI-skilled workers | Gains concentrated in top quartile; entry-level roles suppressed | ✓ Real, uneven |
GDP Impact Official + true |
~$160B in "true GDP" since 2022 (Goldman estimate); $7T projected over a decade | Officially measured: only $45B added to US GDP; "basically zero" in 2025 | ⚖ Understated |
Energy Electricity & water |
Renewables investment pulled forward; grid modernization accelerated | 448 TWh in 2025; 945 TWh projected by 2030; water use to match 1.3B people | ✗ Major cost |
Jobs Displacement & creation |
170M new roles projected by 2030 (WEF); AI-exposed industries saw 3.9% job growth | 85M jobs at risk; 77,999 AI-linked tech job cuts in H1 2025 alone; entry-level hiring collapsed | ⚖ Net positive, painful transition |
Business ROI Firm-level |
66% of companies already seeing efficiency gains; 159% median ROI in SME AI projects (France study) | OpenAI still deeply unprofitable; hyperscalers spending at 30% of sales — triple historic norms | ✓ Real for adopters |
Inequality Winners vs. losers |
Countries and companies early to adopt capture disproportionate gains | Only 32 countries host AI data centers; 90% of capacity in US + China; widening digital divide | ✗ Concentrated gains |
Follow the Capital
The Money In: $1.2 Trillion and Still Accelerating
The AI buildout has become the largest private infrastructure cycle since the 1990s telecom boom. Five US cloud and AI infrastructure spenders — Microsoft, Amazon, Alphabet, Meta, and Oracle — have tripled their capital spending since 2022. And the acceleration is not slowing down.1
VC
Venture Capital · 2025
AI now dominates all of venture capital
In 2025, AI firms captured 61% of all global venture capital — $258.7 billion out of $427 billion total VC deployed. That share has doubled since 2022, when it stood at 30%. Generative AI alone attracted $35.3 billion in VC, up from $2.8 billion the year ChatGPT launched. US firms captured 75% of all global AI VC, with the San Francisco Bay Area alone accounting for more than three-quarters of that.2
$258.7B AI VC in 2025
61% of all venture capital
$700–750B
Hyperscaler Capex · 2026
2026 planned spend: near $700 billion
The five largest US cloud and AI infrastructure spenders — Microsoft, Amazon, Alphabet, Meta, and Oracle (the latter included for its data center capex, though it trails the cloud giants in scale) — have collectively committed approximately $700–750 billion to capital expenditure in 2026 — up from initial guidance of $660–690B as Q1 2026 earnings raised estimates — nearly doubling 2025 levels. Amazon alone projects $200 billion; Alphabet $175–185 billion; Meta $115–135 billion; Microsoft over $120 billion. Goldman Sachs analysis notes that AI capex has consistently come in 50%+ above initial consensus estimates for two years running.3
Amazon: ~$200B
Alphabet: ~$180B
Meta: ~$125B
$1T
The Trajectory · 2027+
Big Tech AI capex is on track to top $1 trillion in 2027
Evercore and Bank of America both project 2027 hyperscaler capex exceeding $1 trillion after Q1 2026 earnings calls. Goldman Sachs' baseline model implies $765 billion in annual AI capex in 2026, growing to $1.6 trillion annually by 2031. The comparison to prior technology booms: AI capex currently represents 0.8% of US GDP, still below the 1.5% peak of the 1990s telecom cycle — suggesting the buildout may have further to run.4
$1.6T projected by 2031
0.8% of GDP today
✦ The Investor's Paradox
The five companies spending the most on AI — Microsoft, Amazon, Alphabet, Meta, Oracle — are simultaneously the most profitable tech companies in history, funding this entirely from free cash flow. This is different from the dot-com era. But OpenAI, the company that sparked the cycle, is still deeply unprofitable and burning capital at a rate that makes it, by some estimates, the largest capital-consuming private company ever. The infrastructure is being paid for by the cash machines. The intelligence layer is still betting on future returns.
What AI Has Delivered
The Gains: Productivity, GDP, and Sector-by-Sector
The productivity story is real — but more nuanced than the headlines suggest. Gains are measurable, uneven, and still early. The GDP numbers depend almost entirely on which measurement methodology you trust. The sector wins are concentrated in a handful of industries.
Productivity — The Real Data
10% lift in AI-exposed industries; 1.1% potential US productivity gain
Industries with higher generative AI exposure from 2017–2024 saw a 10% productivity increase, 3.9% job growth, and 4.8% wage growth per standard deviation of AI exposure (Gallup Workforce Panel, 2026). The Federal Reserve Bank of St. Louis estimated a potential 1.1% productivity increase from generative AI use in the US by late 2024. Firms predict AI will boost their own productivity by 1.4% over the next three years on average — with the gains expected to accelerate, not plateau.5
GDP — The Measurement Problem
"Basically zero" officially. $160B in "true GDP" since 2022.
Goldman Sachs chief economist Jan Hatzius stated in February 2026 that AI added "basically zero" to officially measured US GDP growth in 2025 — due to the import-heavy nature of AI hardware, which means spending doesn't count toward domestic output. A separate Goldman research note estimates AI has added approximately $160 billion to "true GDP" since 2022 — real economic value that the BEA's measurement framework doesn't fully capture. The long-run projections are far more optimistic: Goldman projects $7 trillion in global GDP over a decade; McKinsey estimates $2.6–4.4 trillion annually at full deployment; PwC puts the 2030 contribution at $15.7 trillion.6
Sector Winners — Where the Gains Are Concentrated
Finance, coding, customer service, and healthcare are the early beneficiaries — but the gap between leaders and laggards is growing
According to Deloitte, 66% of organizations are already seeing measurable productivity and efficiency gains from AI — with 45% of companies using generative AI reporting at least a doubling of employee productivity in specific functions. Finance and insurance saw the fastest acceleration in AI-driven productivity since 2022. In software development, AI coding tools have demonstrably compressed certain development cycles. Customer service automation has moved furthest fastest: chatbots and virtual agents now handle interactions that would require thousands of additional human agents. Healthcare and biotech gains are substantial but harder to measure near-term — molecule screening, diagnostics, and clinical trial acceleration are real but have longer commercial payback cycles. French SME data (2022–2025) suggests a median ROI of 159% on AI projects, with payback in 6.7 months on average.7
What AI Has Taken
The Costs: Energy, Water, Jobs, and Capital Burn
The cost side of the AI ledger is not hypothetical — it is already measurable and, in several dimensions, striking. The energy and water numbers are the most visceral. The job displacement numbers are real but more nuanced than the headlines suggest. The capital burn is, for now, historically unprecedented.
⚡
Energy Cost · Electricity
AI data centers consumed more electricity than Saudi Arabia in 2025
Global data centers consumed an estimated 448–490 terawatt-hours of electricity in 2025 — ranking them as the world's 11th largest electricity consumer if treated as a nation. By 2030, AI data centers alone are projected to consume 945 TWh, nearly triple the combined annual electricity use of Pakistan, Bangladesh, and Nigeria. In the US, data centers' total energy demand is expected to nearly double from 80 to 150 gigawatts between 2025 and 2028. Big Tech capex on AI infrastructure reached $448 billion in 2025 — the largest infra spend since the 2000s telecom buildout — most of it directed at power-hungry compute clusters.8
448 TWh in 2025
945 TWh by 2030
80→150 GW in US by 2028
💧
Environmental Cost · Water
AI's water footprint will equal the annual needs of all of Sub-Saharan Africa by 2030
A June 2026 United Nations University report found that AI data centers' associated water footprint by 2030 will equal the basic annual domestic water needs of all 1.3 billion people in Sub-Saharan Africa. Large data centers can consume up to 5 million gallons per day for cooling. In Texas alone, data center water use is projected to reach approximately 49 billion gallons in 2025, rising to an estimated 399 billion gallons in 2030 (Houston Advanced Research Center / Lincoln Institute). Training GPT-3 evaporated an estimated 700,000 liters of fresh water. Only 32% of data center operators even track their water consumption. The land footprint by 2030 will exceed 14,500 square kilometers — roughly twice the Jakarta metropolitan area.9
5M gal/day per large data center
399B gal projected in Texas by 2030
👥
Labor Cost · Jobs
77,999 AI-linked tech job cuts in H1 2025. But the bigger story is hiring suppression, not mass layoffs.
The most important labor market signal from AI so far is not mass displacement — it is suppressed hiring. Goldman Sachs notes that AI appears to be allowing employers to grow without adding headcount, rather than immediately replacing existing workers. In H1 2025, 77,999 tech jobs were directly attributed to AI-driven cuts. AI was cited in 4.5% of all job losses reported in 2025. Software engineering job postings fell ~49% from pre-pandemic baselines. Employment for developers aged 22–25 fell nearly 20% since 2024. The WEF projects 92 million roles displaced but 170 million created by 2030 — a net gain of 78 million — but the transition costs disproportionately hit younger and entry-level workers who cannot find career-starting positions.10
~78K AI-attributed cuts in H1 2025
Software postings down 49%
Net +78M jobs by 2030 (WEF)
✦ The Number Nobody Is Talking About
In Ireland, data centers already account for 21% of total metered electricity — more than all urban households combined. Ireland's national grid operator has paused new data center approvals around Dublin until 2028. In Querétaro, Mexico, fast-tracked data centers are threatening water supplies during prolonged droughts. The communities bearing the physical costs of the AI buildout are not, by and large, the same communities capturing the economic gains.
The Net
The Verdict: Is the World Ahead or Behind?
The honest answer is: it depends on timeframe, geography, and which ledger you're reading. Here's how the categories stack up right now — and what the trajectory looks like.
AI's Net Impact Score — Category by Category
Qualitative assessment based on available data as of Q2 2026. Scale: Net Negative ← → Net Positive. Pointer position represents current state; trajectory arrows indicate direction of travel.
Productivity
+
Real, measurable. Uneven. Still early. Trajectory: accelerating.
GDP (Official)
~
Officially near zero so far. Measurement problem, not economic problem. Trajectory: positive over 3–5 years.
Energy / Climate
−
Clear net cost in the near term. Renewables pull-forward is a partial offset. Trajectory: worsening before it improves.
Jobs
~
Net positive projected by 2030. Transition pain is real and concentrated. Trajectory: painful now, positive medium-term.
Business ROI
+
Strong for early adopters. Still negative for the model builders themselves. Trajectory: broadening.
Inequality
−
AI is widening the gap between first-movers and everyone else — countries, companies, and workers. Trajectory: worsening.
The Bull Case
We are in the infrastructure phase — the productivity payoff comes next
Every major general-purpose technology — electricity, the internet, PCs — followed this exact pattern: enormous capital deployment with delayed, diffuse productivity gains. Penn Wharton projects AI will increase productivity and GDP by 1.5% by 2035, 3% by 2055. The IMF estimates 0.5% annual output lift through 2030. The reason the GDP numbers look small today is that AI is still embedded in intermediate inputs, not final products — a measurement lag, not an economic failure.
The Bear Case
$700 billion a year with "basically zero" GDP return is not a measurement problem — it's a question
Goldman's "basically zero" finding for 2025 is not just a measurement gap — the import-heavy nature of AI hardware (chips, servers) means a large fraction of hyperscaler spending directly accrues to Taiwan, South Korea, and other hardware nations, not the US. OpenAI — the catalyst for this entire cycle — remains the largest capital-burning private company in history. The question is whether the application layer will generate enough revenue to justify the infrastructure layer's spend. That question is still open.
The Honest Take
AI is delivering exactly what every transformative technology delivered in its infrastructure phase: concentrated early gains, distributed future promise, and transition costs borne by those least positioned to absorb them
The gains are real where adoption is deep — in software development, financial modeling, customer service, and AI-skilled workers' wages. The costs are real where the infrastructure lands — in local electricity grids, water tables, and entry-level job markets. The macro payoff is likely real too, but it operates on a 5–10 year lag from investment, not a 1-year lag. The world is not behind. It is in the awkward, expensive middle of a genuinely transformative transition — and the honest scorecard looks exactly like that.
Free · 5 Questions · 2 Minutes
AI Impact Quiz: Who's Actually Winning Where You Are?
The macro numbers tell one story. Your personal exposure to AI's gains and costs depends on your industry, role, geography, and how early you've adopted. Take this quiz to identify which side of the AI ledger you're on — and what to do about it.
AI Impact Quiz
Which Side of the Ledger Are You On?
Five questions. Honest answers. Your result maps to one of four profiles — each with a different set of priorities for the next 12 months.
0 / 5
1 of 5 — Which best describes your current industry?
2 of 5 — How much of your day-to-day work involves tasks that AI can currently do?
3 of 5 — How actively are you using AI tools in your work today?
4 of 5 — How has AI affected your organization's hiring decisions in the past 12 months?
5 of 5 — Thinking about the next 24 months, what is your biggest concern about AI's impact on your situation?
Your AI Impact Profile
Calculating…
Common Questions
Frequently Asked Questions
The five largest US hyperscalers (Microsoft, Amazon, Alphabet, Meta, Oracle) have collectively spent over $1.2 trillion on AI-related capital expenditure from 2022 through 2025, rising from $162 billion in 2022 to $448 billion in 2025. In 2026, their combined planned capex has reached $700–750 billion as of Q1 2026 earnings, revised upward from initial guidance of $660–690B. Global VC investment in AI firms reached $258.7 billion in 2025 alone — 61% of all venture capital deployed worldwide, doubling its 2022 share.
Measured gains are real but modest and uneven. Industries with higher AI exposure from 2017–2024 saw 10% productivity growth. The Federal Reserve Bank of St. Louis estimated a potential 1.1% US productivity increase by late 2024. Goldman Sachs found no meaningful relationship between AI and economy-wide productivity — but identified 30%+ gains in two specific use cases (coding and customer service). The long-run projections are significantly more optimistic: Goldman projects $7 trillion in global GDP over a decade; McKinsey estimates $2.6–4.4 trillion annually at full deployment.
Global data centers consumed an estimated 448–490 terawatt-hours in 2025 — more than Saudi Arabia. They rank as the world's 11th largest electricity consumer. By 2030, AI data centers alone are projected to consume 945 TWh, nearly triple the combined electricity use of Pakistan, Bangladesh, and Nigeria. In the US, data center energy demand is expected to nearly double from 80 to 150 gigawatts between 2025 and 2028. A January 2026 Bloom Energy report calls this adding the energy needs of Spain in just three years.
The headline displacement numbers are smaller than most expect — around 76,000–78,000 AI-attributed job cuts in 2025. But the more important signal is suppressed hiring: AI is allowing companies to grow without adding headcount. Software engineering job postings fell ~49% from pre-pandemic baselines. Developer employment for workers aged 22–25 fell nearly 20% since 2024. The WEF projects 92 million roles displaced globally by 2030 but 170 million new roles created — a net gain of 78 million. Goldman Sachs projects unemployment effects will be transitory and no larger than 0.5 percentage points above trend.
The honest answer is: we're in the awkward middle. The near-term costs are concentrated and visible — energy consumption, job displacement in specific segments, $1.2 trillion in capital not yet generating proportional GDP returns. The gains are diffuse and partially unmeasured. Over the medium term (2027–2035), the projections from Goldman Sachs, McKinsey, the IMF, and Penn Wharton are significantly positive. The transition period — where costs are real and gains are still accruing — is where we are now. This is not unusual. It is the pattern of every major general-purpose technology in its infrastructure phase.
Disclaimer: This article is for informational purposes only and does not constitute financial, investment, or career advice. All data points are sourced from publicly available research as cited. Figures marked "approximate" or "projected" reflect estimates from third-party sources and are subject to revision — all 2026 capex and projection figures are forward-looking and will be updated in Q3 2026 as actuals become available. AI economic projections are speculative by nature and should not be used as the sole basis for any decision.
References
Sources & References
1
Epoch AI / Visual Capitalist (April 2026). Big Tech AI Spending Over Time (2022–2025). visualcapitalist.com ↗
2
OECD (2026). Venture Capital Investments in Artificial Intelligence through 2025. oecd.org ↗
3
Futurum (February 2026). AI Capex 2026: The $690B Infrastructure Sprint. Updated Q1 2026 earnings raised combined Big Four estimate to ~$725B (Financial Times/Tom's Hardware, April 2026). futurumgroup.com ↗; Goldman Sachs (December 2025). Why AI Companies May Invest More than $500 Billion in 2026. goldmansachs.com ↗
4
CNBC (April 2026). AI boom: Big Tech capital expenditures now seen topping $1 trillion in 2027. cnbc.com ↗; Goldman Sachs (May 2026). Tracking Trillions: The Assumptions Shaping the Scale of the AI Build-Out. goldmansachs.com ↗
5
Gallup Workforce Panel / Phys.org (April 2026). Industries most exposed to AI are not only seeing productivity gains but jobs and wage growth too. phys.org ↗; St. Louis Fed (October 2025). Generative AI, Productivity and the Future of Work. stlouisfed.org ↗; CEPR / Yotzov et al. (March 2026). Firms predict an AI productivity boom is coming. cepr.org ↗
6
Goldman Sachs / Fortune (September 2025). AI has added $160 billion to 'true GDP' since 2022. fortune.com ↗; Gizmodo (February 2026). AI Added 'Basically Zero' to US Economic Growth Last Year, Goldman Sachs Says. gizmodo.com ↗
7
Orange / Deloitte (May 2026). Artificial intelligence in business: productivity and governance in 2026. orange.com ↗
8
United Nations University / UNU-INWEH (June 2026). Environmental Cost of AI's Energy Use: Carbon, Water and Land Footprints. unu.edu ↗; Consumer Reports (March 2026). AI Data Centers: Big Tech's Impact on Electric Bills, Water, and More. consumerreports.org ↗
9
Time / UNU-INWEH (June 2026). AI Could Use as Much Water as 1.3 Billion People by 2030, U.N. Report Warns. time.com ↗; Lincoln Institute (February 2026). Data Drain: The Land and Water Impacts of the AI Boom. lincolninst.edu ↗
10
World Economic Forum. Future of Jobs Report 2025. Stanford HAI. 2026 AI Index. Indeed Hiring Lab (2026). AI Skills in Job Postings. hiringlab.org ↗; World Data (March 2026). AI Job Displacement Statistics 2026. theworlddata.com ↗
Not financial advice. This article is for informational and educational purposes only. It does not constitute financial, tax, investment, or legal advice. Always consult a qualified professional before making financial decisions. Full disclaimer →