Google just dropped its biggest AI push yet at I/O, with a new Gemini Omni model, a 24/7 agent called Spark, and a price cut on its top tier. Bitcoin is grinding sideways around $77,000 with traders openly debating whether February's $60,000 low was the bottom or whether $70,000 is next. Babylon's Bitcoin staking protocol just crossed $4 billion in TVL, and Aave V4 is moving to integrate native BTC lending. Trump signed an executive order pushing the Fed to open payment rails to crypto firms. And the EU just delayed its toughest AI Act deadlines into 2027 and 2028. Let's get into it.
Google I/O 2026 was unmistakably an AI keynote, and the numbers Sundar Pichai opened with tell you why. The Gemini app went from 400 million monthly users at last year's I/O to 900 million today. Daily requests are up sevenfold. AI Overviews in Search now hit 2.5 billion users a month. That is not a side project anymore. That is the product.
The headline release is Gemini 3.5 Flash, the new default model, plus Gemini Omni, which Google is calling a world model. Omni takes text, images, and video as inputs and generates video as output, with character identity preserved across scenes. It's rolling into the Gemini app, Google Flow, and YouTube Shorts. If you're a creator, the bridge between prompt and finished clip just got a lot shorter.
But the more interesting piece is Gemini Spark. This is Google's general-purpose agent. It runs in the cloud, 24/7, even when your device is off. It's wired into Gmail, Docs, Slides, and Calendar. It can monitor your inbox for recurring charges, turn a stack of notes into a polished brief, kick off projects on its own, and ask permission before high-stakes actions. There's also a Daily Brief feature that aggregates your email, calendar, and chat history into a morning digest with suggested next steps.
On pricing, Google cut its top AI Ultra tier from $250 to $200 a month, same capabilities. And they're shifting away from daily prompt limits to a compute-based usage model that refreshes every 5 hours up to a weekly cap. Hit the ceiling on big models and you get auto-routed to smaller ones, with pay-as-you-go top-ups available.
The competitive read is straightforward. OpenAI and Anthropic have been pulling ahead on the agent narrative for months. Google just answered with deeper Workspace integration, a billion-user distribution channel, and a price cut. If Spark actually works as advertised, the bar for what counts as an AI assistant just moved.
Bitcoin has been stuck in a tight range around $77,000 for three days now, down more than 4% on the week. The picture underneath is more interesting than the price.
On the bearish side, Bitcoin failed again to reclaim its 200-day moving average near $82,000. US bond yields are pushing near 20-year highs. Bitcoin ETFs just saw over $1 billion in outflows, the first negative week in seven. Short-term holders dumped roughly 10,000 BTC at a loss, around $770 million worth. Analysts at Wintermute and others are flagging $76,000 as the line in the sand, with $70,000 to $74,000 as the next downside zone if it breaks.
But there are real arguments the bottom is in. CoinDesk highlighted three metrics from on-chain analysts: realized cap is stabilizing, RHODL readings are at historically elevated levels associated with cycle lows, and funding rates are deeply negative, meaning traders are paying to be short. That combination has marked bottoms before.
Research firm K33 made a related point: this bear market is different because traders are uniquely pessimistic. Leverage is low. Open interest on futures is down. That actually limits downside, because there's no crowded long position to cascade-liquidate.
And then there's the bullish camp. One price model is targeting $255,000 by year-end as a conservative case. Bernstein and Arthur Hayes are both calling for new highs in 2026. On Bitfinex, margin longs just hit a two-and-a-half year high. Traders are buying this dip aggressively.
Meanwhile, implied volatility on Bitcoin options is staying low even as price falls and yields rise. Options desks are recommending long straddles, betting on a big move in either direction. The setup is coiled. Somebody is going to be very wrong.
Here's a milestone worth paying attention to. Babylon, the Bitcoin staking protocol, just crossed $4 billion in total value locked, one year after launch. What makes Babylon different from every other attempt at Bitcoin DeFi is the structural choice: no wrapping, no bridging. Bitcoin stays on Bitcoin. The staking is verifiable on the native chain.
That matters because wrapped BTC and cross-chain bridges have been the source of nearly every major Bitcoin DeFi blow-up. Babylon's pitch is that you can make Bitcoin productive without giving up the security model that makes it Bitcoin in the first place.
The next step is bigger than the staking number. Aave V4, which just launched its hub-and-spoke architecture on Ethereum mainnet, has a proposal in governance to integrate Babylon for native BTC lending. The idea: deposit native Bitcoin as collateral, borrow USDC or USDT, without ever wrapping the BTC or handing it to a custodian. The mechanism uses zero-knowledge proofs to verify the state of Bitcoin from Ethereum.
And here's the cost detail that makes this newly feasible. On-chain ZK verification used to cost thousands of dollars per proof. It's now in the $10 to $20 range. That's the kind of cost collapse that flips theoretical from practical.
If this works, the implications are real. Bitcoin's market cap dwarfs everything else in crypto, but the vast majority of it just sits there. A trustless lending market on native BTC turns trillions of dollars of dormant collateral into productive capital, without forcing holders to trade away custody or take bridge risk.
There's a Bitcoin-maximalist caveat, of course. Plenty of holders will never want their coins anywhere near a smart contract or a staking commitment. That's fine. But for the segment of the market that does want yield and leverage on Bitcoin terms, the infrastructure is finally arriving. The next test is whether Aave's DAO actually ships the integration and whether the loans get used.
Two regulatory moves on opposite sides of the Atlantic, both worth tracking.
First, Trump signed an executive order on May 19 directing the Federal Reserve to review how crypto and fintech firms get access to the central bank's payment systems. This is the Fed master account question, the same fight Custodia Bank lost in court. The order asks the Fed to evaluate whether depository institutions, including crypto-adjacent ones, can be granted access to payment services. It also directs regulators to streamline applications for bank and credit union charters for eligible fintech firms.
The names that benefit if this goes through are exactly who you'd expect: Kraken, Ripple, Coinbase, and Circle. Direct Fed rails would cut out correspondent banks, lower costs, and make stablecoin issuance more competitive against the traditional banking stack. Senator Warren has already started pulling on the thread, questioning the OCC chief about crypto trust charter approvals and possible links to the White House. Expect a fight.
On the other side, the EU just blinked, sort of. The provisional Digital Omnibus deal reached on May 7 delays the toughest enforcement deadlines for the AI Act. Stand-alone high-risk AI systems now have until December 2027. AI used as a safety component in products has until August 2028. Watermarking obligations got pushed to December 2026. The rationale is that the technical standards aren't ready, and you can't enforce rules that don't have implementation specs.
Meanwhile in the US, there are roughly 1,200 AI bills in circulation across federal and state levels, with no unified standard for evaluating any of them. The White House is reportedly considering an executive order modeled on FDA-style approvals, vetting frontier AI models before release. That would be a sharp turn from the deregulatory posture earlier in the administration.
The through-line: governments everywhere are realizing that AI policy is downstream of national security, financial stability, and labor market disruption all at once. No single agency has the expertise, and the timelines keep slipping. For companies actually deploying this stuff, the practical move is to build documentation and governance now, because the rules will land eventually and they'll land harder than the delays suggest.
One prediction to close on. If Babylon and Aave actually ship native Bitcoin lending without bridges or wrapped tokens, that will matter more in five years than anything Google announced this week. Agents come and go. Custody-free Bitcoin collateral changes the shape of the entire market.