Is Bitcoin a safe haven?
In recent weeks, the Bitcoin market has behaved noticeably differently from many traditional markets. Some see this as evidence that the currency could play a role as a geopolitical safe haven. After that, we’ll look at a new application in the crypto world: training AI models through decentralized networks. Find out more in the market update below.
Market update
The relative strength of Bitcoin and many other cryptocurrencies over the past two weeks has surprised many observers. Since the start of the war on 28 February, the price of Bitcoin has risen by 15%, while the major US stock index, the S&P 500, has fallen by 3%, and gold even dropped by 5%.
Some analysts see this as proof that Bitcoin functions as a safe haven and performs well during periods of geopolitical turmoil. Like gold, Bitcoin has a certain degree of neutrality and independence. No one is in charge, no single party can change the rules on their own, and no one can print more at will.
It is an appealing explanation, but we should approach it with caution. A similar narrative circulated in February and March 2022, during the first weeks of the war in Ukraine. A few months later, the crypto market plunged further into the bear market.
A more likely explanation is that we’re simply seeing a temporary rally within a cyclical bear market, similar to the rise to €84,000 in January. Only a weekly close above the 50-week average at €85,000 would provide the first indication that we are at the start of a new bull market.
Featured
Crypto network trains AI model
Many of the AI applications we use every day run on so-called large language models (LLMs). Think of chatbots and AI agents that help with programming or writing. These models are created by processing enormous amounts of data and learning patterns. Training typically takes place in large data centers of tech companies such as OpenAI, Google, or Meta, where thousands of powerful machines work together under one roof.
A decentralized alternative to the tech giants is now emerging. This week, the Templar project announced that it has completed the largest decentralized training of a language model to date. The model, Covenant-72B, contains 72 billion parameters and was trained on over 1.1 trillion tokens. Instead of relying on a single data center, the team used the Bittensor network, a protocol in which participants provide computing power and are rewarded with the TAO token.¹
The training lasted about six months, from September 2025 to March 2026. On average, seventeen participants took part in each round, with a total of more than seventy unique nodes deploying their hardware across the world.
That might sound like a network of hobbyist computers, but that isn’t the case. Participants ran heavy-duty infrastructure: each node used at least eight professional GPUs and about 1.2 terabytes of RAM. That represents a hardware investment of several hundred thousand euros per participant.
The real technical breakthrough lies elsewhere. The main challenge in decentralized AI training is communication. In a data center, all hardware is connected via extremely fast networks. Over the public internet, those connections are much slower. Templar therefore developed a method to drastically reduce the amount of data exchanged between participants. As a result, a communication round took only about seventy seconds, instead of the several minutes required in earlier experiments.
The result is a model that, according to early benchmarks, can compete with some existing open-source models trained in traditional data centers. One striking difference is the training data: Covenant-72B was trained using roughly half the number of tokens.
For the crypto market, the underlying network is particularly interesting. Bittensor aims to build a marketplace for AI computing power, where participants are paid for their contributions. The success of this training run was rewarded by investors: TAO rose by about one-third in value this week, to around 270 dollars.²
Sources:
In other news
Bitcoin network proves surprisingly resistant to cable outages. Research from Cambridge shows that even if 72% to 92% of all international subsea internet cables fail at the same time, the network would largely continue to function. Historic data on cable breaks show that 87% of such disruptions affect less than 5% of the nodes. Targeted attacks on critical infrastructure pose a more serious risk. According to the study, the use of the Tor network also contributes to the system’s resilience.
Alibaba invests in a Singapore-based stablecoin and payments company. The Chinese e-commerce giant is investing 35 million dollars in MetaComp, a Singapore company that develops infrastructure for using traditional currencies and stablecoins for international payments and wealth management. MetaComp processed more than 10 billion dollars in transactions last year and is already operating profitably. The company plans to expand across Asia, the Middle East, Africa, and Latin America.
BlackRock launches ether fund with built-in staking yields. On Wednesday, the asset manager listed the iShares Staked Ethereum Trust (ETHB). The fund buys ETH and then locks it up for staking. Investors receive roughly 80% of the staking rewards each month. On the first trading day, 43.5 million dollars in new capital flowed into the fund. Total assets briefly topped 100 million dollars.
Tether investing in infrastructure for Bitcoin payments. In total, Ark Labs has raised 7.7 million dollars from investors, including Tim Draper and Anchorage Digital. The company is working on a new payment layer for Bitcoin. Ark bundles transactions so that users can pay instantly without using complex payment channels, as used in the Lightning Network. For Tether, the goal is clear: bring USDT back onto the Bitcoin network.
Sources:
Satoshi Radio: The latest episode of Satoshi Radio looks at countries such as Kazakhstan and South Africa, which are interested in bitcoin mining. Nasdaq wants to accelerate its tokenization initiatives. The hosts also cover US regulators who made peace, the payment vision of the Dutch central bank, and a critical report on the Dutch approach to anti-money laundering. As always, the market update concludes this episode.
This article is for informational purposes only and does not constitute a marketing communication or recommendation. None of the content herein should be considered as investment advice or a substitute for it. Bitvavo makes no guarantees regarding the accuracy or completeness of the provided information. Investments involve risks. There is a possibility of losing your entire invested capital.