Crypto's AI Utility Questioned as Xiaomi Hits 1,000 TPS and Tokenized RWAs Surge 589%

Crypto News

Academic researchers from the Initiative for CryptoCurrencies and Contracts argue that crypto has only limited usefulness in solving artificial intelligence's deepest problems around trust and payments. In a survey published Monday, the group pushed back on the popular claim that handing AI agents a crypto wallet makes them autonomous, writing that a wallet enables automation but does not make a model smarter or harder to shut down. The authors also dismissed the idea that a blockchain can reliably separate human content from machine-generated material, noting it is well-suited to timestamping artifacts but not to verifying authorship at scale.

Xiaomi, better known for smartphones and scooters, broke a major AI inference record on Monday with MiMo-V2.5-Pro-UltraSpeed, a serving mode for its trillion-parameter flagship that exceeds 1,000 tokens per second and peaks near 1,200. Working with inference partner TileRT, the company achieved the milestone on a standard eight-GPU commodity node rather than custom silicon. For context, mainstream chat models typically run between 68 and 192 tokens per second. The speed comes from FP4 quantization on the model's expert layers and a speculative decoding technique that proposes full blocks of tokens at once. A limited API trial runs June 9 through June 23 at triple standard rates.

Tokenized real-world assets remain one of the few growth sectors in an otherwise cooling market, with active tokenized RWAs climbing 589% from early 2025 to June 2026. Bonds and money market funds led in dollar terms, adding $6.5 billion, while tokenized stocks posted the fastest growth at 422%. Tokenized precious metals gained $1.5 billion as geopolitical uncertainty fueled safe-haven demand earlier in the year. The expansion comes despite a sharp early-June selloff in Bitcoin and broader markets, driven by rate-hike expectations, uncertainty over the CLARITY market structure bill, and Strategy's sale of 32 BTC.

Apple used its annual developer conference on Monday to unveil Siri AI, a rebuilt voice assistant capable of holding conversations, analyzing on-screen content, drawing on personal context, and completing multi-step tasks across devices. It marks the biggest overhaul of the assistant since its 2011 debut and follows a troubled Apple Intelligence rollout that forced feature delays and a class-action lawsuit over marketing claims. Apple Intelligence is also expanding into Safari, Messages, Mail, Calendar, Photos, Home, and Shortcuts. The announcement came during Tim Cook's final keynote as chief executive before John Ternus assumes the role on September 1, with the new Siri set to launch in beta later this year.

The debate over autonomous AI agents intensified as several firms pushed deeper into agent-controlled finance. MetaMask launched a non-custodial wallet designed specifically for AI agents, with Consensys founder Joe Lubin arguing that machine intelligences will increasingly transact and coordinate on crypto rails built for autonomous actors. Robinhood, meanwhile, signaled that customers will soon be able to deploy AI agents to trade crypto on their behalf, separate from their main portfolios, starting with an equities beta. Researchers caution that such agents remain dependent on human-built infrastructure, and that conventional payment systems can already handle much of the same automation without on-chain access.

Institutional and retail tokenization efforts accelerated alongside the RWA boom. Tokenized SpaceX shares drew fresh attention after Kraken began offering a tokenized equivalent of the private company's stock through the xStocks platform, which surpassed $25 billion in cumulative trading volume within about eight months. Ondo Global Markets crossed $1 billion in total value locked in the same window. In real estate, Apex Group started providing fund services on Goldman Sachs' Digital Asset Platform, while banks increasingly explore tokenized deposit networks to modernize payments and compete with the rapid rise of stablecoins, signaling that on-chain settlement is moving from DeFi experiments toward core financial plumbing.

Taken together, these developments sketch a market in transition rather than retreat. Even amid a near-term bear market driven by rate fears and regulatory ambiguity, the dominant narrative this cycle is convergence: AI and blockchain are colliding around payments, autonomous agents, and tokenized assets, with traditional institutions—from banks to exchanges—accelerating adoption of on-chain rails. Yet the IC3 findings inject a sober counterweight, reminding the industry that infrastructure speed and tokenized growth do not automatically resolve questions of trust, oversight, and genuine utility. The lasting story is institutional maturation tempered by hard scrutiny of what crypto can—and cannot—actually deliver for AI.