Billionaire Jeremy Grantham: Bitcoin won’t suddenly go to zero, but will quietly fade away.

Binance founder Changpeng Zhao (CZ) said there is no single cause for the crypto market’s sharp downturn in the first half of 2026. Geopolitical tensions, investors shifting funds to AI, and the typical four-year crypto cycle may have collectively driven the continued decline of Bitcoin and other crypto assets. Bitcoin hit an all-time high of over $126,000 last October, and has since fallen roughly 50%. It opened near $89,000 at the start of this year, briefly rose to just over $96,000, then dropped to around $60,000. In the long term, the crypto industry will keep growing, with rising demand for fintech as transaction volumes continue to increase, so he is not worried about the sector itself or short-term price fluctuations. He noted that emerging sectors like AI are absorbing "hot money" from crypto, though this could be a positive factor in the long run. On prediction markets, CZ said they are growing rapidly as tools for price discovery and liquidity, which is beneficial for the public. Regarding regulation, CZ said standalone bills like the U.S. Digital Asset Market Clarity Act are important but tactical matters that will not determine the crypto industry’s long-term growth. He hopes the Clarity Act passes, warning that if U.S. related legislation is delayed, other countries may take the lead in setting rules. CZ also stated that if U.S. Democrats regain control of at least one chamber of Congress after the midterm elections, there could be scrutiny of Trump’s support for the crypto industry and his pardons of crypto executives. He emphasized he has "nothing to hide" and is willing to cooperate if parties seek information. On political implications, CZ said he tries to stay away from U.S. politics, but believes any anti-crypto figures could lose a significant number of votes now.

According to Axios, citing sources, the Trump administration is nearing approval to allow Anthropic to restore access to its powerful Fable 5 model, which has been offline for 15 days due to government security concerns. Insiders expect the administration could lift restrictions on Fable 5 as early as next week. Another source said relevant communications are set to continue over the weekend, with Anthropic poised to resume access to Fable soon. On Friday, the U.S. Department of Commerce permitted Anthropic to restore access to Mythos 5 for a limited number of trusted users. Per Semafor, Commerce Secretary Lutnick wrote in a Friday letter to Anthropic that the company "has collaborated with the U.S. government to address risks related to Mythos 5 and Fable 5." "These efforts have made significant progress." Additionally, Anthropic has committed to cooperating with the U.S. government on agreements, standards, and releases. (Jinshi)

Strategy (MSTR) has seen its modified net asset value (mNAV) fall below 1, indicating the market is currently valuing the company at less than the worth of its Bitcoin holdings. This is unusual for Michael Saylor-led Strategy. For years, investors have priced Strategy at a premium to its Bitcoin reserves, giving the company flexible access to capital when needed—a advantage Saylor and his team have leveraged heavily. Currently, Strategy’s share price has dropped to around $82, roughly 85% lower than its November 2024 all-time high, bringing its enterprise value to approximately $50.4 billion. Meanwhile, with Bitcoin trading at about $60,000, the value of Strategy’s Bitcoin holdings stands at roughly $51.1 billion. That means the market is now valuing the entire company at less than the value of its Bitcoin assets. At this valuation level, issuing new shares would be dilutive for Strategy, as the company would effectively sell equity at a discount to its underlying asset value. While this does not bar Strategy from continuing to issue new shares, raising capital at current valuations could spark more criticism. The firm’s recent Bitcoin purchases have already diluted common shareholders and drawn community backlash. Market concerns have grown that Strategy is increasingly resembling a closed-end fund rather than an operating company. Such vehicles typically trade at a premium to their underlying Bitcoin holdings when demand is strong, but can trade at persistent discounts once investor sentiment weakens. However, unlike traditional closed-end trusts, Strategy still retains multiple tools: issuing debt or equity when it is accretive, redeeming or refinancing securities, generating operating cash flow through its software business, and actively managing its capital structure.

According to Coinglass data, Coinbase’s Bitcoin Premium Index has remained in negative premium for 40 consecutive days, currently standing at -0.1569%, reflecting sustained weak purchasing power in the U.S. market. The Coinbase Bitcoin Premium Index measures the gap between Bitcoin prices on Coinbase and the global average market price. A negative premium typically signals heavy selling pressure in the U.S. market, declining investor risk appetite, rising market risk aversion, or capital outflows.

According to data from Farside Investors, U.S. spot Bitcoin ETFs posted a combined net outflow of $1.7873 billion this week. Among them, BlackRock’s IBIT saw a net outflow of $1.3035 billion, Fidelity’s FBTC recorded a net outflow of $314.9 million, and Grayscale’s GBTC had a net outflow of $135.3 million. Meanwhile, some ETFs registered net inflows: BTC ETF saw a net inflow of $71.7 million, and MSBT posted a net inflow of $26.2 million.

Coinbase CEO Brian Armstrong published a post stating that to sustain exponential growth in token usage while keeping AI spending stable, the key is not to introduce usage friction or spending alerts, but rather better default models, routing, and caching mechanisms. Coinbase is testing using open-weight models like GLM 5.2 and Kimi 2.7 as defaults via its LLM gateway, while still encouraging engineers to select the right model for each task. He noted that 91% of employees have never hit their usage caps, so instead of lowering quotas and adding alerts, the company shifted to lower-cost default models. For model routing, Coinbase preprocesses prompts in its custom workflow and routes tasks to the most suitable model based on cache hit rates and model pricing. For example, the planning phase may require an advanced model, but using an advanced model during execution would be overkill. The company believes that in the future, humans should not choose models—AI should handle this task automatically. Armstrong also said that cache misses are the easiest way to drive up costs. All of Coinbase’s requests are cache-aware to reuse hot caches as much as possible; for instance, after proper cache implementation, LibreChat’s cache hit rate rose from 5% to 60%. Additionally, Coinbase requires engineers to keep contexts streamlined, including starting new sessions when switching tasks, narrowing file context ranges, and disconnecting unused tools. The goal is not to curb AI usage, but to build infrastructure that can support exponential growth. Through these practices, Coinbase has cut its AI spending by nearly half, while token usage continues to grow.