AI and crypto: the quiet revolution making some rich and firing others

This story doesn't have a single hero. The same technological shift is simultaneously generating double-digit returns for traders, pushing certain protocols' valuations into the billions — and has already cost the crypto industry more than 5,700 jobs in just the first half of 2026. Here's how artificial intelligence actually works inside the crypto market today — no hype, no doom, just the facts.

Where AI Is Genuinely Being Used in Crypto Projects

The label "AI token" can hide anything from a genuine product to a simple rebrand chasing a trendy narrative. But a handful of projects have actually built working infrastructure at the intersection of AI and blockchain, rather than just adding the word "intelligence" to their pitch.

• Bittensor (TAO) — an "incentive layer" network for distributed compute: miners earn the TAO token by providing real computational power for specific AI tasks inside specialized "subnets" — for example, training large models. By 2026, the network runs more than 100 subnets, and the project's market cap has reached $3.2-3.4 billion, making Bittensor the largest project by market cap at the intersection of crypto and AI
• Render Network (RNDR) — a decentralized rendering and compute network that has already onboarded NVIDIA's Blackwell (B200) architecture to offer enterprise-grade compute; part of the protocol's revenue goes toward buying back and burning the token, tying its price directly to real compute demand
• Fetch.ai — after merging with SingularityNET and Ocean Protocol into a single ASI token, the project has focused on autonomous economic agents — programs capable of independently making deals and executing tasks on-chain without human involvement
• Virtuals Protocol — a no-code platform for creating and launching AI agents, each with its own token and potential revenue from apps, games, and DeFi protocols; the platform has launched more than 40,000 tokenized agents and is expanding across several chains, including Arbitrum, XRP Ledger, and BNB Chain

An important caveat: even genuinely working projects have highly volatile tokens. In March 2026, RENDER and TAO both jumped 35-40% on positive industry news, while Virtuals Protocol's VIRTUAL token peaked at $5.07 before correcting sharply to under $1 — even as the underlying product and the platform's expansion to new chains kept moving forward uninterrupted. Having working technology doesn't insulate a token from speculative swings, and that's worth keeping separate from the question of whether the product actually works.

Where AI Is Actually Turning a Profit Right Now

The most mature and measurable application of AI in crypto today isn't tokens with "AI" in the name — it's automating trading itself. Industry reviews estimate that experienced traders running well-configured bots achieve net annual returns 5-25% above simply holding the asset, while institutional AI-driven scalping strategies target daily returns of around 2.5%. An important caveat: the result depends far more on how the bot is configured, which market it operates in, and whether the trader understands what they're doing — than on the AI itself. The algorithm alone doesn't print money.

We've already covered a real example with real numbers on CryptoPulse.media: in our piece on how trading bots earned a private trader $1,150 in a month, the headline figure was +86.75% in a month at 10x leverage — a vivid illustration of both the upside and the elevated risk of that kind of strategy.

At the institutional level, the picture is more serious: crypto-focused hedge fund Numerai Signals manages roughly $200 million in assets and has already paid out about $50 million to its top performers via models that predict market moves. Large trading desks run proprietary AI models for high-frequency trading across multiple exchanges at once, capturing arbitrage opportunities that close in under a second — a window no human could physically react to in time.

Another Growth Area: AI Security Auditing

AI's profitability in crypto isn't just about trading. The second most mature application is smart contract security auditing. A traditional manual audit costs anywhere from $50,000 to $500,000 and can mean waiting up to 12 weeks in an audit firm's queue. AI tools like Anthropic's Mythos and Coinbase's Frosty can scan a contract for vulnerabilities in minutes — Anthropic estimates the average cost of such a scan at around $1.22 per contract.

Even more significant than one-off checks is continuous AI monitoring of already-deployed protocols: industry reviews show that projects using ongoing monitoring cut incident-related losses by more than 80%. CertiK, the largest Web3 security provider, combines AI with formal verification to protect and monitor protocols.

The approach has a clear ceiling, though: human auditors still catch what AI misses — economic-incentive attacks, social engineering, and complex business logic that doesn't match a recognizable pattern. The industry consensus is that AI scanning as a first-pass filter, followed by targeted human review, is the new baseline — not a full replacement for auditors.

The Other Side: Mass AI-Driven Layoffs

Alongside rising profitability in algorithmic trading, the crypto industry is going through a wave of layoffs that companies are directly attributing to AI-driven automation. Industry tallies show more than 30 crypto companies cut a combined 5,700-plus positions in the first half of 2026 alone.

• Crypto.com cut about 12% of its staff — roughly 180 of 1,500 employees — while integrating AI into internal processes
• Coinbase announced layoffs affecting about 14% of its total workforce
• BitGo cut roughly 85 positions — about 15% of staff — framing it as a refocus on security, trading, stablecoins, settlement, and "AI-powered infrastructure"
• Kraken cut about 150 roles as AI tools expanded across the exchange
• Block (Jack Dorsey's company, which includes Cash App and its bitcoin business unit) cut about 4,000 employees — nearly half its staff, down from over 10,000 to under 6,000

Not Just Cuts: How Coinbase Is Restructuring

Coinbase's case shows that the "14% of staff" headline hides a full operating-model overhaul, not just a budget trim. Rob Witoff, Coinbase's head of platform, said close to 100% of the company's code is now written by, or with the direct involvement of, large language models — up from roughly 40% as recently as February of the same year. In parallel, CEO Brian Armstrong eliminated the role of "pure managers," replacing them with "player-coaches":

“Managers should be like player-coaches, getting their hands dirty alongside their teams. Layers slow things down and create coordination tax.”

— Brian Armstrong, CEO of Coinbase, X post, May 2026

Quote source: Fortune

After the restructuring, Coinbase's management hierarchy is capped at no more than five layers below the CEO himself. Even companies cutting headcount keep hiring selectively, though: Kraken, for instance, cut 150 roles while simultaneously keeping dozens of positions open, concentrated specifically in compliance and engineering — the exact areas AI can't yet fully replace, only augment.

Dorsey directly tied the decision to AI-driven efficiency gains, saying he wasn't willing to wait until the change became unavoidable:

“The intelligence tools we're creating and using, paired with smaller and flatter teams, are enabling a new way of working which fundamentally changes what it means to build and run a company. I'd rather get there honestly and on our own terms than be forced into it reactively. Within the next year, I believe the majority of companies will reach the same conclusion and make similar structural changes.”

— Jack Dorsey, CEO of Block, February 2026

Quote source: Fortune

It's also telling just how seriously major players are betting on AI as a strategy, not just a cost-cutting measure: Crypto.com CEO Kris Marszalek paid a record $70 million for the ai.com domain in early 2026, framing it as a decades-long bet:

“Taking a 10 to 20 years view, AI represents one of the greatest technological waves of our lifetime. There is a big desire for us to own this touchpoint, otherwise you get commoditised.”

— Kris Marszalek, CEO of Crypto.com

Quote source: CoinDesk

The Next Wave: Agents and the "Machine Economy"

Market maker Wintermute describes an emerging trend that hasn't fully played out yet: crypto becoming a settlement layer not for people, but for autonomous AI agents that pay each other for services and data. A deeper dive into this — including the x402 protocol and a real Wintermute quote — is available in our separate piece on crypto's next growth wave. A parallel, physical layer of automation is developing too: Norwegian company 1X Technologies, for instance, unveiled 25-degree-of-freedom, tactile-sensing hands for its NEO humanoid robot — a sign that the AI-and-real-world convergence goes well beyond trading algorithms.

Tellingly, even Wintermute itself acknowledges that over the past year, AI investment has been pulling capital away from crypto that might otherwise have flowed into it — meaning that until recently, the two sectors have been competing for the same institutional investors more than reinforcing each other. The "machine economy" thesis describes a shift from that competition toward technical convergence, but by Wintermute's own admission, that shift is still far from complete.

A Small Forecast: Where This Is All Heading

If the recent months' trend holds, value in the AI-crypto pairing will keep shifting away from tokens with a trendy "AI" prefix and toward infrastructure that actually processes transactions — stablecoins, protocols like x402, and low-fee networks. On the jobs side, the pressure is unlikely to ease: operational, compliance, and support roles will likely bear the brunt, while demand for trading-algorithm specialists, AI-system security experts, and agent developers themselves will probably keep growing. One indicator worth watching isn't the price of specific "AI tokens," but the actual volume of machine-to-machine transactions moving through agentic payment protocols — that's what will show whether the "machine economy" has become reality rather than just a thesis.

This material is for informational purposes only and is not investment advice.