Bitcoin Holds Near $64K as AI-Led Risk-Off Deepens Extreme Fear

Crypto News

The artificial-intelligence trade cracked this week after China's Moonshot AI released Kimi K3, a model our desk sees reshaping the competitive map. The system carries 2.8 trillion parameters and a one-million-token context window, making it the world's largest open-source release. On the frontend code arena leaderboard it scored 1,679, edging past the strongest closed models from Anthropic and OpenAI to claim the top spot. Backed earlier by a $1 billion round from investors including Alibaba, Moonshot has since seen its valuation climb to roughly $31.5 billion. The result triggered what traders are already calling the Kimi shock across equity desks worldwide.

The reaction in chip stocks was immediate and severe. The Philadelphia Semiconductor Index plunged about 10% in a single week and slid more than 20% from its June peak, formally entering bear-market territory. Heavyweights including Nvidia and Broadcom led the decline as investors questioned whether the astronomical capital committed to AI infrastructure can still earn a return. The damage rippled across Asia: Chinese rival Z.AI fell 28% and MiniMax dropped 16%, while SoftBank, a major OpenAI backer, sank 9%. Semiconductor names in Korea, Taiwan and Japan were dragged lower in tandem, underscoring how concentrated and crowded the global AI hardware trade had become.

A second cheap-model catalyst deepened the rout. Z.ai unveiled GLM-5.2, an open-source system able to handle up to one million tokens of context and engineered to cut the compute cost of running advanced AI. The release revived the DeepSeek-style question that haunted markets last year: if software becomes dramatically more efficient, does the industry still need so many GPUs and memory chips? Our reading is more nuanced. The Jevons paradox suggests cheaper compute tends to expand total usage, as firms move from a single chatbot to hundreds of AI agents and automated systems such as an AI trading bot. Efficiency lowers cost, yet it also manufactures demand.

Compounding the technology-driven selloff, geopolitical risk flared at the worst possible moment. The United States expanded large-scale strikes on Iran, and Tehran threatened neighboring energy shipping routes, collapsing weeks of fragile ceasefire hopes. Markets that had treated a pause in fighting as peace were reminded that a ceasefire and a lasting settlement are not the same thing. Crude oil prices surged on fears of disrupted Middle East supply, adding an inflationary shock to an already jittery tape. The overlap of a geopolitical flare-up with an AI valuation reset left global risk assets, from equities to digital assets, with few obvious havens this week.

Beneath the headlines, a large rotation was already under way. For the past year, AI hardware had absorbed nearly every available dollar; this week capital visibly rotated out of that trade and into financials, energy and healthcare, along with sectors that had lagged the rally. The key question our desk is weighing is whether the AI industry itself is finished or whether AI equities simply climbed too far, too fast and are now correcting. Those are very different diagnoses. History offers a caution: when DeepSeek first appeared, many declared AI infrastructure spending over, yet global data-center investment subsequently grew larger, not smaller.

Korea absorbed the sharpest blow because its market carried both heavy AI exposure and heavy leverage. Samsung Electronics and SK Hynix dominate the Kospi, so when they fall together the whole index struggles to hold. On top of that, retail investors had piled into margin trading and single-stock leveraged products. During the plunge, an estimated 1.2 million leveraged accounts faced additional collateral demands and roughly 350,000 were forcibly liquidated. Forced selling of that scale does not unwind in an orderly way: margin calls do not ask about next quarter's memory demand, they simply sell today, and that pressure can spread to Japanese, Taiwanese and US hardware names.

Tying these threads together, this week was a synchronized deleveraging across AI equities, semiconductors and commodities, and crypto sat squarely in its path. Our proprietary market data captures the mood: the Fear and Greed Index has collapsed to 25, deep in Extreme Fear, while total crypto market capitalization stands near $1.84 trillion. Bitcoin (BTC) is holding close to $64,000 and Ether (ETH) near $1,844 as of this writing, with Bitcoin dominance elevated at 69.8% — a classic sign that capital is huddling in the largest asset and fleeing the altcoin market. Until forced selling in traditional risk assets exhausts itself, digital assets, including AI-linked tokens and AI crypto wallet projects, are likely to trade as high-beta proxies, far from their all-time high.