โ† Back to HomeBack to Blog List

Anthropic Researcher Says: AI Will Surpass All Humans in a Few Years, $4 Trillion Spending by 2028

๐Ÿ“Œ Key Takeaway:

Anthropic's RL lead predicts AI will surpass all humans within years, capable of doing everything humans do on computers. Global AI spending may hit $4 trillion by 2028.

Anthropic's Key Researcher: AI Will Surpass All Humans Within Years โ€” Global AI Spending to Hit $4 Trillion by 2028

๐Ÿ“Œ Key Takeaway: Sholto Douglas, Anthropic's head of reinforcement learning, predicts AI that surpasses all humans could emerge within the next few years โ€” capable of handling any work humans do on computers. Global annual AI capital expenditure may reach $4 trillion by 2028. Manus closed a $500M+ round at a $4B valuation, with Tencent and Sequoia doubling down. Capital is flooding into the AI Agent race.

"Surpass all humans" โ€” those words carry different weight when spoken by someone building AI safety.

Sholto Douglas leads reinforcement learning at Anthropic. Anthropic โ€” the company that builds both the most powerful AI models (Claude) and researches "how to keep AI from destroying humanity." AI safety is literally their core mission.

When someone like this says, publicly, "AI that surpasses all human capabilities could appear within a few years," that's not marketing. That's an assessment from the absolute frontier of AI research.

What Exactly Did He Say?

Douglas's core claim is clear: an AI system capable of performing all work that humans do on computers could emerge within years.

Note the keyword: all.

Not "some work." Not "most work." All of it. Writing code, running analysis, doing design, managing projects, writing reports, conducting research โ€” any knowledge work done on a computer, AI will be able to handle independently.

His team at Anthropic is developing the next generation of Claude models. Claude is already one of the strongest AI models available. And their internal assessment of the next generation's capabilities points directly to this "surpass all humans" conclusion.

The Insane $4 Trillion Number

More mind-blowing than the technical predictions are the money numbers.

Industry forecasts show that by 2028, global annual AI capital expenditure will reach $4 trillion. To put that in perspective โ€” that's roughly Vietnam's entire annual GDP. Every single year.

Where's the money going? Training larger models, building more data centers, buying more chips, deploying more AI agents. Microsoft, Google, Amazon, Meta, Apple โ€” every tech giant is pouring money in at unprecedented speed.

NVIDIA's market cap has blown past $3 trillion, making it the most valuable company on Earth. CEO Jensen Huang says "AI factories will replace traditional factories." That's not bragging โ€” the capital expenditure plans from every major tech company prove they actually believe it.

Manus Raises $500M, Tencent and Sequoia Keep Betting

Zoom out from the giants and look at startups โ€” the scene is equally wild.

AI Agent startup Manus just closed a funding round exceeding $500 million at a $4 billion valuation. The investor lineup is stacked โ€” Tencent, Sequoia, and other top-tier firms keep doubling down.

What does Manus build? General-purpose AI Agents โ€” systems that let AI autonomously handle complex tasks. This is exactly the "capable of all computer work" AI that Douglas described above, turned into a real product.

Why is capital going this crazy? Because everyone sees the same signal: the commercialization inflection point for AI Agents has arrived.

In 2024, people were debating "what is an AI Agent?" In 2025, they're debating "who can build the best AI Agent?" By 2026, that question may already have become "who's using it the most?"

What Does This Mean for Everyone Else?

Here's the honest take.

When "AI surpasses all humans" shifts from prophecy to plan, when trillions of dollars flood into the sector, the optimal strategy for ordinary people isn't panic โ€” it's adaptation.

Learn to use AI tools โ€” not just surface-level skills like "writing prompts," but genuinely learning to use AI to accomplish things you previously couldn't. If one person's output can be amplified 10x by AI, you won't be replaced by AI.

Douglas said "within years." Maybe 3 years, maybe 5. But not much longer. Get ready.

Want Better SEO Results?

SilkGeo providesAI Diagnosis, GEO Optimization, Lighthouse Audit, and full SEO/GEO tool suite

Use SilkGeo for free