Bitcoin spiked above 72,000 dollars after a surprise US-Iran ceasefire, triggering 600 million dollars in liquidations and pulling roughly 3 billion dollars into Binance alone. Morgan Stanley launched its spot Bitcoin ETF today with a market-low 0.14 percent fee, directly challenging BlackRock's 55 billion dollar IBIT fund. Google dropped Gemma 4, an open-source model family where a 31 billion parameter model is outperforming rivals ten times its size. And MARA Holdings sold over 15,000 Bitcoin to retire a billion dollars in debt, slashed its workforce, and is pivoting hard into AI infrastructure. Let's get into it.
Let's start with the macro picture, because the last 24 hours have been genuinely wild. Late Tuesday, President Trump confirmed a Pakistan-brokered two-week ceasefire with Iran, tied to reopening the Strait of Hormuz. Oil crashed below 100 dollars a barrel. Bitcoin surged to 72,734 before pulling back to around 71,500. Risk assets across the board caught a bid, while volatility indexes dropped. What's interesting is that Bitcoin actually started moving before the ceasefire announcement went public. CryptoSlate reported that the breakout past 71,000 happened in after-hours trading, as markets began unwinding the fear that Iran's deadline would spiral into a broader energy shock. Once Trump posted, it was confirmation of a move already underway. 600 million dollars in short liquidations followed. On-chain data tells a compelling story underneath this. CoinDesk reported that nearly 850,000 Bitcoin changed hands in the 60,000 to 70,000 dollar range during the recent dip — an enormous accumulation zone. And according to Cointelegraph, over 4.37 million BTC now sit in long-term holder wallets, with a network activity index flipping to what analysts are calling a bull phase signal. But perhaps the most structurally significant news today is Morgan Stanley's spot Bitcoin ETF going live. The Morgan Stanley Bitcoin Trust, ticker MSBT, launches with a 0.14 percent expense ratio — the cheapest on the market. That undercuts BlackRock's IBIT, which currently dominates with roughly 55 billion in assets. Morgan Stanley isn't just competing on fees though. They're plugging into a 7 trillion dollar wealth management engine. If even a small fraction of those advisory accounts rotate in, it reshapes ETF flows meaningfully. JPMorgan noted that digital asset inflows totaled just 11 billion dollars in Q1, with Strategy — Michael Saylor's company — practically the only entity keeping institutional crypto flows alive. Morgan Stanley entering the picture could change that math. Still, the ceasefire is only two weeks, and leverage remains elevated. So bulls got a reprieve, but the caution is warranted. Oh, and one more thing on the geopolitical front — Iran is reportedly weighing a plan to charge oil tankers a Bitcoin-denominated toll for safe passage through the Strait of Hormuz. One dollar per barrel of oil, paid in Bitcoin. If that's real, it's not just a headline — it's a sovereign state using Bitcoin as sanctions-resistant trade infrastructure. That's a use case nobody had on their bingo card.
Now let's talk about what's happening beneath the surface of Bitcoin's network — the mining layer — because the numbers there paint a very different picture from the rally on your screen. Bitcoin's global hashrate dropped 5.8 percent in Q1 2026, falling from around 1,066 exahashes per second down to 1,004. That's the sharpest quarterly contraction in recent memory. And it's gotten worse since — Bitcoinist reports the network is now hovering around 953 exahashes. The cause is straightforward. Bitcoin's price fell from roughly 124,000 dollars late last year to the mid-60s in early 2026. That cratered miner revenues. Hashprice — the dollar value per petahash per second per day — hit a record low near 27.89 dollars. For context, the average cost to produce one Bitcoin now exceeds 80,000 dollars, while the market price sits around 69,000 to 72,000. Many miners are operating at negative margins. The result: about 252 exahashes of older, less efficient hardware has been shut down. A 10 percent difficulty adjustment followed, trying to stabilize the economics for remaining miners. And the pain isn't evenly distributed. Iran's hashrate collapsed 77 percent over the past quarter, largely due to the ongoing conflict, though analysts say profitability is the bigger driver even there. Meanwhile, the US still leads globally with 37.4 percent of hashrate, followed by Russia and China — together those three countries control 65 percent of Bitcoin's mining power. That concentration is worth sitting with. Bitcoin's security model depends on decentralization, and having two-thirds of mining power in just three nations creates a persistent vulnerability, even if no single actor controls a majority. There's a silver lining in the data though. Emerging markets like Kyrgyzstan and Paraguay are growing rapidly, deploying newer, more energy-efficient equipment. And March brought a symbolic milestone — the 20 millionth Bitcoin was mined, meaning roughly 95.2 percent of all Bitcoin that will ever exist is now in circulation. With only about a million coins left to mine over the next century, the economics of mining will increasingly depend on transaction fees rather than block rewards. The halving in April 2024 already cut rewards in half, and the stress we're seeing now is a direct consequence. The broader trend among large miners is diversification. Several major operators are pivoting toward AI data centers, citing higher and more stable margins. Which brings us neatly to the next segment.
MARA Holdings — the company formerly known as Marathon Digital — just made one of the most dramatic moves we've seen from a public Bitcoin miner. They sold 15,133 Bitcoin for approximately 1.1 billion dollars. That's not a trim. That's a strategic liquidation. With the proceeds, MARA repurchased roughly 1 billion dollars of its convertible debt at a 9 percent discount, reducing total liabilities by about 30 percent. They also cut 15 percent of their workforce — around 40 positions — and transferred another 250 Bitcoin, worth about 17 million dollars, to new wallet addresses afterward. Their remaining Bitcoin treasury sits at 38,689 BTC. Shares are down about 22 percent over the past year, trading near 8.80 dollars. And MARA's leadership is clearly signaling that pure-play Bitcoin mining is no longer the path forward. The company is investing in AI data center leasing, partnering with firms like Exaion and Starwood to generate more stable enterprise revenue from its existing energy and infrastructure assets. This is not unique to MARA. It's becoming the playbook for large-scale miners post-halving. Mining rewards got cut in half in April 2024, Bitcoin's price has been volatile, energy costs remain high, and the hashprice is at record lows. The math simply doesn't work for many operators at current prices unless you have the newest hardware and the cheapest power. So the pivot makes sense: these companies already own or lease massive amounts of power capacity and physical infrastructure. Repurposing that for AI workloads — where margins are better and demand is exploding — is a rational move. But it does raise a question for Bitcoin maximalists. If the largest miners are systematically de-risking away from Bitcoin, what does that mean for network security over time? Especially as hashrate is already declining and geographic concentration is increasing. The bull case is that this shakeout is healthy. Weaker miners exit, difficulty adjusts downward, and the remaining operators become more profitable and more resilient. That's how Bitcoin has always worked through cycles. The bear case is that this cycle's mining economics are structurally different — with production costs above market price for many operators, the industry can't simply wait it out. Meanwhile, a new bill from Senators Lummis and Cassidy — the Mined in America Act — proposes a federal certification program for US-based mining facilities and a direct procurement channel through the Treasury to buy newly mined Bitcoin at a fixed price. If that passes, it would create a government demand floor for domestically mined Bitcoin and could reshape the entire mining landscape. It's early days for that legislation, but it shows Washington is paying attention to the strategic implications of where and how Bitcoin gets mined.
Let's shift to AI, because Google just dropped something that deserves attention. Gemma 4 is a new family of open-source models ranging from 2 billion to 31 billion parameters, all released under an Apache 2.0 license. And the headline is this: the 31 billion parameter dense model is outperforming rivals with 400 billion parameters on key benchmarks in math, coding, and science. That's not a marginal improvement. That's an order of magnitude more efficient. And it aligns with a broader shift in the industry away from brute-force scaling toward smarter architecture and better training recipes. What makes Gemma 4 notable beyond raw performance is its versatility. Every model in the family is multimodal — handling text, images, video, audio, and OCR. That's an industry first at this scale for open-weight models. The smallest variants can run on a Raspberry Pi or a smartphone. The larger ones fit on a single consumer GPU. Context windows go up to 256,000 tokens. They support function calling, configurable thinking modes, and native system prompts. For enterprises, this is significant. It means you can self-host a genuinely capable AI model without renting cluster time from a hyperscaler. The architecture isn't revolutionary — Google DeepMind focused on reinforcement learning refinements, data quality, and training optimization rather than fundamental architectural innovation. But the results speak for themselves. This connects to a bigger theme in AI right now. An article from the AI Journal argues that brute-force scaling — just adding more parameters, more data, more compute — is hitting diminishing returns. The cost curves are unsustainable. Future progress is increasingly coming from compression, pruning, knowledge distillation, and targeted optimization. Gemma 4 is a proof point for that thesis. The competitive implications are real. Anthropic recently reported surging revenue from API and enterprise services — they're clearly winning on the proprietary side. But Google releasing models this capable under a permissive open license puts pressure on everyone. If a 31 billion parameter model you can run locally competes with proprietary giants, the moat for closed-source AI gets thinner. For the Bitcoin and crypto world specifically, this matters because cheaper, more capable AI running on modest hardware accelerates everything from on-chain analytics to autonomous agents to vulnerability detection. Speaking of which — Anthropic's Claude Mythos Preview reportedly identified thousands of zero-day vulnerabilities across major operating systems, browsers, and cryptography libraries that DeFi infrastructure depends on. AI models have reached a level where they rival or exceed most human experts at finding and exploiting software flaws. That's a double-edged sword, and it's arriving fast.
Here's the thing to sit with today. Bitcoin miners are selling their Bitcoin to buy AI infrastructure. Google is proving that smaller, smarter models beat brute-force giants. And a sovereign state may start collecting Bitcoin tolls on the world's most important oil chokepoint. The common thread is efficiency displacing scale — in mining, in AI, in geopolitics. The entities that survive this cycle won't be the biggest. They'll be the leanest. That's it for today. I'll talk to you tomorrow.