AI News Today: 3 moves toward a 10 trillion dollar future from Stripe, Anthropic, and Perplexity

AI’s next $10 trillion opportunity is less about “AI as software” and more about AI-powered services that make customers dramatically faster—if companies can crack the unit economics behind payments, search, and agent workflows. A Sequoia-style framing shifts the addressable market from roughly a $0.5T software-only pie to a $1T+ software-and-services pie, arguing that AI’s effectiveness in services expands where the money can be made and why ROI matters more than ever.

Three developments from the last day illustrate how that larger market could be captured. First, Stripe’s rumored $1.1 billion acquisition of Bridge looks like a classic “boring but decisive” infrastructure move. Bridge sits in crypto and stablecoins, offering APIs for stablecoin payments. The strategic logic is hard-headed: card networks like Visa and Mastercard take a non-trivial cut—often 1% to 3%—on Stripe’s enormous transaction volume (over $1 trillion annually). Stablecoins could let merchants and users shift some payment flows away from card rails, enabling frictionless payments and unlocking use cases that are difficult with traditional card economics.

Micro-payments are the clearest example. Online payments often struggle with a lower pricing floor (commonly around $5), which makes tiny transactions uneconomic. Stablecoin rails could enable “micro” interactions—especially important in AI-driven products where users may pay for many small steps. Bridge also points toward a wallet-style experience: moving funds into stablecoins inside a closed loop managed by Stripe could reduce or eliminate transaction fees, with savings shared across Stripe, merchants, and customers. Stripe’s history of being cautious about payment mechanics makes the acquisition notable; it signals a belief in a real path to ROI rather than a purely speculative crypto bet.

Second, Anthropic’s push into agentic search and agentic reasoning—tied to the release of Claude 3.5 Sonnet—signals a more experimental phase of AI tooling. Anthropic is training models for agentic tasks where the system can drive the computer and complete work with less step-by-step micromanagement. The tradeoff is cost and reliability: the agentic “computer use” approach is token-heavy and can get distracted or stuck. Still, Anthropic positions these capabilities for developers first, even as consumer imagination gravitates toward scenarios like ordering lunch via an agent—raising obvious safety and liability concerns.

Third, Perplexity is taking similar underlying agent/search technology and narrowing it into a more immediately useful product: extended search where an agent can independently browse the web and answer questions. Wrapped into Perplexity Pro, the scope is constrained—search instead of full mouse-and-keyboard control—which makes it easier to deploy and iterate. That constraint also accelerates value discovery: Perplexity can find “use cases inside existing AI software” faster than systems that must manage broader computer interactions.

Taken together, the outlook suggests three layers of winners in a $10 trillion services economy: infrastructure plays (Stripe/Bridge) that enable the payment rails AI needs; narrow, high-ROI application plays (Perplexity’s search); and later-stage, more speculative general-purpose agent capabilities from major model labs (Anthropic), which may take time to become broadly useful. The common thread is services value: the biggest gains come when specific workflows get solved dramatically faster—and when the underlying costs, from token pricing to transaction fees, fall enough to scale.