NVIDIA. xAI. Anthropic. OpenAI. Perplexity.

All of them are racing to build the AI agent layer that dominates.

NVIDIA's version is called NemoClaw.

Enterprise-grade. Open source. Secure.

Jensen Huang is reportedly designing it first for businesses, then for consumers—and committing $26 billion over five years to back that up.

Think about what that means.

NVIDIA started at the bottom of the AI stack—the chips. Then moved up to the models that run on the chips. Now they're moving up to the agents that run on the models that run on the chips.

That's the Apple playbook.

Own the hardware. Own the software. Own the application layer on top of both.

xAI and Tesla are trying something similar. They’ve just launched Digital Optimus—an AI that lives on your computer, moves your mouse, does your job. Runs on a $650 chip already inside millions of cars.

The race isn't about who has the best model anymore.

It's about who owns the agent layer that sits on top of all of them.

We’re going to talk about this—and a lot more—during our Tech Turning Point 2026 event, live in San Jose.

(If you haven’t yet signed up to get updates, we’re nearing the last call.)

For now, though, let’s dive a little deeper.

To understand where this is all headed, let’s see what’s already being done across industries with AI agents:

What’s Already Happening

BNY Mellon runs 134 “digital employees” on its Eliza platform, working 24/7 across contract review, trade settlement, anomaly detection, and compliance. Contract review time dropped 75%—from four hours to one. That platform now supports over 125 live use cases, with 20,000 employees actively building agents.

BlackRock published research on "AlphaAgents," a multi-agent system where three specialized AI agents (fundamental, sentiment, and quantitative analysis) debate each other to construct equity portfolios—mimicking an investment committee. In backtesting, the multi-agent portfolio outperformed both single-agent and equal-weighted benchmarks on returns and Sharpe ratios.

Goldman Sachs became the first major bank to deploy Cognition's Devin autonomous coding agent across its 12,000-person engineering team, reporting 3-4x productivity gains compared to the 20% from GitHub Copilot. Goldman also deployed Anthropic's Claude for trade accounting and client onboarding, where agents now review documents, extract entities, and make "micro decisions" in KYC workflows.

Morgan Stanley has 100,000+ proprietary research reports—decades of institutional knowledge. Before their AI assistants, advisors were accessing 20% of it. The rest was buried. Manual searches took 30 minutes per query.

Now they access 80% of the library. Each advisor saves up to 15 hours a week. Adoption: 98%.The firm is now building what wealth management head Jed Finn calls "super agents" composed of "hundreds, if not thousands" of smaller AI agents—and these super agents will "eventually" assemble investment portfolios.

 Wells Fargo used an LLM-driven agent system to re-underwrite 15 years of loan documents and is embedding AI agents into call centers and branch operations.

Walmart built what it calls four "super agents" spanning customer service, fashion production, developer tools, and personalized shopping. One agent cut 18 weeks off fashion production timelines. Another handles customer support with no human handoff.

Reddit deployed Salesforce Agentforce and deflected 46% of support cases. Resolution time dropped from 8.9 minutes to 1.4 minutes. Advertiser satisfaction up 20%.

Solopreneurs: the Underrated Disruption

The most underappreciated AI story isn't happening in major institutions.

It's happening in spare bedrooms.

Alan Wells raised $6.5 million from Y Combinator to build Rocketable—a company that acquires profitable SaaS businesses and replaces the entire team with AI agents. One person. Full operations. Customer support, development, marketing. All agents.

Nat Eliason built an AI agent called Felix—"CEO of a zero-human company." Felix built a website, integrated Stripe, deployed sub-agents for support and sales. Revenue: $195,000 in five weeks.

The tools making this possible are now absurdly cheap.

Meanwhile, solo-founded startups jumped from 23.7% to 36.3% of all new startups between 2019 and mid-2025. 58% of small businesses now use generative AI.

Tech CEOs have a standing bet on the first year a one-person billion-dollar company emerges.

Dario Amodei says it could happen in 2026.

He’s probably right.

The Hype is Real, BUT…

The pattern across all three user groups—Fortune 500, Wall Street, solo operators—points to the same conclusion: AI agents crossed from experimental to operational in 2025, and 2026 is the scaling year.

The companies reporting hard dollar figures (JPMorgan's $1.5-2B in value, TD Bank's $170M, BofA's 11,000-employee-equivalent savings, Cardinal Health's write-off reduction from $20M to $35K) are beyond running pilots.

But there are potholes.

McKinsey found only 10% of organizations have scaled AI agents across any individual function. And Klarna's dramatic supposed AI-driven workforce cuts led to quality problems that forced re-hiring. PwC reported that the vast majority of organizations say current AI spending hasn't produced measurable business returns yet.

Forbes columnist Gene Marks argues most small businesses are still just "monkeying around with a chatbot."

And yet…

Every major player this week—NVIDIA, Tesla, Meta, Google, Perplexity—is signaling one thing:

The agent layer is the prize.

And the war is just getting started.