History Is Rhyming
The EV industry made a fatal assumption — that if you build enough cars, charging infrastructure, and government incentives, adoption will follow. It didn’t, cleanly. BloombergNEF reduced its EV adoption forecasts for the first time in 2025, largely due to policy rollbacks and sluggish consumer confidence. EV makers who poured billions into gigafactories are now sitting on overcapacity. Sound familiar?
AI in 2026 is running the exact same playbook — and heading toward the same reckoning. Hyperscalers are expected to spend $400B+ on GPUs and data centers, while AI revenue growth lags dangerously behind. According to M&G Investments’ analysis, capex is accelerating while revenue stalls — a divergence that compounds with every hardware cycle.
Act 1: The GPU Investors — Building an Empire on Sand?
The first victims in a low-adoption world won’t be the non-adopters. They’ll be the ones who bet everything on adoption happening fast.
Goldman Sachs, SoftBank, OpenAI, Oracle, and MGX have committed $500 billion over four years in US data centers alone. Google plans to spend $75 billion on AI infrastructure in 2025 alone. Sundar Pichai himself warned internally that “the risk of underinvestment is considerably high” — but here’s the uncomfortable truth: only a fraction of the existing $65 billion GPU market in 2024 was even being used for AI applications.
When adoption lags infrastructure buildout, you get the telecom bubble of 1999 — fiber laid that no one used, companies that went bankrupt holding assets that took a decade to become useful. The AI version: GPU farms sitting at 30% utilization, while hyperscalers race each other into overcapacity because no single player can afford to be the one who blinks first.
The EV parallel here is brutal: Battery manufacturers who scaled for 30% YoY EV growth hit a wall when growth outside China clocked only 6.1%. The GPU investor’s nightmare is precisely this scenario — massive sunk costs, rapid architectural obsolescence (Nvidia ships new GPU generations every 18 months), and enterprise customers who are still in “pilot mode”.
Act 2: The Cloud Bill Payers — Death by a Thousand Invoices
The second casualty is the enterprise that started adopting AI — but only halfway.
According to a Tangoe survey of 500 IT and finance professionals, 72% say AI-themed cloud spending is becoming unmanageable, with costs rising 30% in 2024 alone. MIT’s GenAI Divide study (2025) is even more damning: 95% of companies that invested $35–40 billion collectively in GenAI saw zero measurable returns. The reason? Brittle workflows, no contextual learning, and AI deployed in areas — sales and marketing — where human judgment still dominates.indianexpress+1
This is the EV equivalent of someone who bought an electric car but lives 40 km from the nearest charger. They paid the premium, they’re stuck with the inconvenience, and they tell everyone the technology doesn’t work. The problem isn’t the technology — it’s the missing ecosystem.
For enterprises, the missing ecosystem is: change management, clean data pipelines, AI-literate workforces, and clearly defined use cases. Without these, cloud bills pile up, pilots fail, and boards lose faith in the entire AI thesis. 83% of AI leaders in 2026 report major or extreme concern about GenAI — and that’s among the people who are already using it.
The dangerous middle ground: Half-adopted AI is worse than no AI. You pay the costs, absorb the disruption, but capture none of the competitive advantage.
Act 3: The Non-Adopters — Comfortable Until They’re Not
Now for the quietest, most dangerous position of all — the companies that watched the EV hype, watched the AI hype, and decided to wait it out. They’re the ones who feel smartest right now.
They shouldn’t.
Gartner projects that 40% of enterprise applications will ship with embedded AI agents by 2026, up from less than 5% in 2025. Early movers are already clocking a 3x revenue-per-employee advantage over non-adopters in AI-exposed sectors. McKinsey data shows 92% of enterprises plan to increase AI investments — which means the non-adopter isn’t just staying still, they’re falling behind an accelerating curve.
The historical precedent is grim and well-documented:
- Kodak invented the digital camera in 1975 — and buried it to protect film revenues. They filed for bankruptcy in 2012.
- Blockbuster turned down Netflix for $50 million in 2000. Netflix is worth over $100 billion today.
- Nokia dismissed the iPhone as “not how phones work” — and lost over $100 billion in market value.
None of these companies were stupid. They were comfortable, and comfort in a tech disruption cycle is the most expensive luxury in business.
The AI version of Kodak is already being written. It’s the mid-size BPO that thinks “our clients won’t trust AI outputs.” It’s the regional bank that’s waiting for “regulatory clarity.” It’s the logistics firm that says “our drivers know the routes better than any algorithm.” By the time they’re ready to adopt, the switching cost won’t be a software license — it’ll be a decade of lost compounding advantage.
The Three-Way Trap: A Matrix of Pain
| Stakeholder | The Bet They Made | The Risk If Adoption Stays Low |
|---|---|---|
| GPU/Infrastructure Investors | Adoption will catch up to supply | Overcapacity, GPU price collapse, telecom-bubble repeat |
| Partial Cloud Adopters | ROI will emerge from pilots | Spiraling cloud bills, no ROI, board-level AI fatigue |
| Non-Adopters | The technology will plateau or fail | 3x competitor productivity gap, Kodak-moment obsolescence |
What Actually Needs to Happen
The EV market’s real problem wasn’t the car — it was the infrastructure gap between the product and the user’s life. Charging networks, range anxiety, resale value uncertainty. AI’s infrastructure gap is cognitive and organizational, not physical. It’s about:
- Workforce AI fluency — teams that know how to prompt, validate, and integrate AI outputs
- Data readiness — clean, structured, accessible data that AI can actually act on
- Use-case discipline — starting with back-office automation (where MIT confirms AI works) before attempting sales, creativity, or judgment-heavy functions
- Leadership commitment — not just a pilot budget, but a transformation mandate
The enterprises that crack this in the next 18 months won’t just win — they’ll become the competitive moat that new entrants and late movers find nearly impossible to breach. In AI, like EVs, the early infrastructure builders will eventually be vindicated — but only if the market meets them. The question is whether Indian enterprises, in particular, will be the ones building that bridge or the ones crossing it after everyone else already has.
“The AI adoption window isn’t closing — but the cost of walking through it late is rising every quarter. The GPU investors need you. The cloud providers need you. But more urgently: your future self needs you to decide now.”
