Thursday, March 26th. A humanoid robot escorts the First Lady at the White House. Morgan Stanley prepares to put its own name on a spot Bitcoin ETF. Bhutan quietly dumps two thirds of its national Bitcoin treasury. And Mistral drops an open-source voice model that runs on a smartwatch. Let's get into it.
Okay, let's start with something you probably didn't expect to see this week. First Lady Melania Trump was escorted by a humanoid robot at a White House summit on Tuesday. The robot, Figure 03, literally introduced itself, saying quote, I'm Figure 03, a humanoid built for the United States of America. And honestly, the robot stole the show. The event was called the Fostering the Future Together Global Coalition Summit, focused on AI and education, but the visual of a walking, talking humanoid next to the First Lady is what everyone's going to remember.
This isn't just a novelty moment. Figure is one of the most well-funded robotics companies in the world, and putting their hardware in the White House is a very deliberate signal about where U.S. policy attention is heading. Robotics and physical AI are becoming a national priority.
Meanwhile, Tesla is not sitting still. The company is aggressively recruiting for its Optimus Gen 3 humanoid, and the design language they're going for is, and I quote, superhero in a suit. They want the thing to look genuinely human, with soft-goods aesthetics and natural proportions. Tesla has development teams spread across Palo Alto for AI, Fremont for mass production, and even sites in Canada and Greece for motor and gear design. The robot already does laundry and cleaning tasks, and they're pushing toward general-purpose utility.
What's interesting about the Tesla approach is the vertical integration story. They've got their own neural network training infrastructure from full self-driving, their own compute hardware, and now they're applying all of that to a physical form factor. The robot walks at up to 5 miles per hour and is being positioned for factory deployment to address labor shortages.
The broader humanoid robot market is projected to hit $38 billion by 2035, and right now we're watching the race take shape between well-funded startups like Figure and mega-corporations like Tesla. Both paths have advantages. Figure can move fast and make splashy demos at the White House. Tesla can manufacture at scale. The next couple of years will tell us which approach wins, or whether there's room for both.
Now let's talk about Bitcoin, because the price action this week is messy but the underlying story is actually more nuanced than the headlines suggest.
Bitcoin slipped below $69,000 on Wednesday as oil prices rebounded on fading Middle East peace hopes. The Iran situation has sent crude surging, bond yields climbing, and risk assets broadly under pressure. The CoinDesk 20 index fell 3.2% with every single constituent trading lower. It was an ugly day across the board.
But here's where it gets interesting. JPMorgan put out a note saying Bitcoin is actually holding ground better than gold and silver, which are both sliding on ETF outflows and liquidity strains. The bank noted that institutional flows into Bitcoin have been steadier, with improving momentum even amid geopolitical stress. That's a meaningful data point. When a major Wall Street bank says Bitcoin is showing more resilience than traditional safe havens during a war scare, that's worth paying attention to.
On the technical side, Bitcoin has now traded in a tight range for nearly 50 days. Some people are calling it a bear flag, but CoinDesk's analysis argues it's structural consolidation rather than a bearish continuation pattern. The supply in profit metric dropped below 50% back in February, and historically, the last time that happened, Bitcoin gained 655% in the cycle that followed. We're currently 44% off the $126,000 all-time high, and multiple on-chain indicators suggest we're in the later stages of this bear market.
The big institutional news is Morgan Stanley. The NYSE posted a listing notice for what would be the first bank-issued spot Bitcoin ETF, ticker MSBT. This is not a fund that Morgan Stanley distributes for someone else. This would be their own product with their own name on it, selling Bitcoin directly to clients. If this launches, it's a landmark moment for institutional adoption.
On the corporate treasury front, Strategy continues to dominate. CryptoQuant data shows Michael Saylor's company accounted for nearly all recent corporate Bitcoin purchases, buying about 45,000 BTC in the last 30 days alone. Other companies' share of treasury purchases has collapsed from 95% down to about 2%. The $100 billion corporate Bitcoin trade is essentially one buyer right now, which is both impressive and a concentration risk worth watching.
Let's shift to a story that's been building quietly but deserves more attention. Bhutan, the tiny Himalayan kingdom, has been aggressively selling its national Bitcoin reserves. And the numbers are striking.
Bhutan's holdings have dropped 66% from a peak of nearly 13,000 BTC in late 2024 down to just 4,453 BTC, worth roughly $315 million at current prices. In 2026 alone, the country has moved over $152 million worth of Bitcoin out the door. The latest transaction was 519 BTC, about $36.75 million, transferred to external addresses on Wednesday.
What's notable is the acceleration. Earlier this year, Bhutan was moving $5 to $15 million per week. In March, that jumped to $35 to $45 million per week. Most of these sales are being conducted through over-the-counter deals, often with Singapore-based QCP Capital, so they're not dumping directly onto exchanges and crashing the price, but the selling pressure is real.
Here's the context that makes this story fascinating from a Bitcoin maximalist perspective. Bhutan didn't buy its Bitcoin. It mined it using hydroelectric power. The country has abundant cheap hydro energy, and the government-backed mining operation was actually a smart way to monetize a natural resource. At the peak, Bhutan was sitting on over $900 million in Bitcoin, making it one of the most concentrated national Bitcoin positions relative to GDP anywhere in the world.
So why sell now? Some analysts suggest portfolio rebalancing rather than outright capitulation. Bhutan may be taking profits to fund development projects. But there's a problem. The country had pledged 10,000 BTC toward the Gelephu Mindfulness City project, and with holdings now at 4,453, that commitment looks increasingly difficult to honor.
This contrasts sharply with El Salvador, which despite all the criticism continues to hold its Bitcoin and even buy the dips. Bhutan's sell-off is a reminder that nation-state Bitcoin adoption isn't a one-way street. Governments face political pressures, budget constraints, and short-term thinking that can override long-term conviction. The lesson for Bitcoiners is that sovereign adoption is great, but it's only as durable as the political will behind it.
Let's close with something from the AI side. French AI company Mistral just released Voxtral TTS, an open-source text-to-speech model, and it's genuinely impressive.
Voxtral supports 9 languages including Hindi and Arabic, and here's the headline feature: it can clone a custom voice from less than 5 seconds of audio. Not 5 minutes. 5 seconds. It captures accents, intonation, rhythm, and emotional tone from that tiny sample and can then generate speech in any of its supported languages. That means you could record a brief English clip and have the model generate fluent Arabic speech in your voice.
The technical specs matter here. It's built on top of Ministral 3B, which is Mistral's small language model, and it hits 90 milliseconds time-to-first-audio with a 6x real-time factor. Translation: it generates speech 6 times faster than you can listen to it. And it's small enough to run on edge devices. We're talking smartphones, smartwatches, not cloud servers.
In human evaluations, Mistral claims Voxtral outperforms ElevenLabs in naturalness, which if true is a big deal because ElevenLabs has been the gold standard in AI voice generation. And unlike ElevenLabs, this model is open source. You can download it, run it locally, modify it, build products on top of it.
The strategic angle here is important. Mistral is positioning itself as the open-source alternative across the entire AI stack: text, code, vision, and now voice. They're going after enterprise use cases like customer support, real-time translation, and dubbing. And they've announced plans for a full multimodal platform that integrates audio, text, and images.
This is part of a broader trend where open-source AI models are closing the gap with proprietary ones at remarkable speed. A year ago, open-source voice models were noticeably worse than commercial ones. Now Mistral is claiming parity or better, and giving it away for free. For anyone building voice-enabled applications, this changes the economics completely.
One quick security note to end on. Separately this week, hackers compromised a popular open-source AI tool called LiteLLM and injected code that searched for crypto wallets and private keys every time Python started up. The malicious versions were live for about 5 hours before being caught. It's a reminder that as open-source AI tools become more critical infrastructure, supply chain security becomes paramount. Always verify your packages.
Here's the thought to sit with today. Morgan Stanley putting its own name on a Bitcoin ETF, Figure putting a robot in the White House, Mistral giving away state-of-the-art voice AI for free. The institutions aren't just watching anymore. They're building, branding, and shipping. The question isn't whether these technologies go mainstream. It's whether you're positioned for it when they do. I'll talk to you tomorrow.