A company that was almost bankrupt just surpassed Bitcoin in market value.

Author: Zhou, ChainCatcher

On June 22, SK Hynix's stock price rose, pushing its market value to $1.35 trillion, surpassing Bitcoin's total market cap of about $1.29 trillion, and during trading, it temporarily overtook Samsung Electronics to become Korea's most valuable company.

According to Coinglass data, in the global asset rankings, SK Hynix rose to 16th place, while Bitcoin slipped to 18th place.

HBM, and a 13-year-long bet

The core driver behind SK Hynix's recent surge is HBM (High Bandwidth Memory). AI training and inference demand extremely high memory bandwidth, and SK Hynix is NVIDIA's main HBM supplier, with a market share exceeding 60%.

Financial reports show that SK Hynix's Q1 revenue was 52.58 trillion won, operating profit was 37.61 trillion won, with a profit margin of 72%. Analysts currently consensus on SK Hynix's Q2 operating profit is around 62–65 trillion won, with some brokerages' optimistic forecasts raised above 68 trillion won.

In early April this year, market expectations for Q2 mostly remained in the 50 trillion won range, but as memory prices continued to stay strong, brokerages generally revised their forecasts upward significantly. During the earnings call, management stated that the structural memory shortage brought by AI will last for several years at least, and plans to significantly increase capital expenditure to expand advanced capacity.

It is reported that SK Hynix has been betting on HBM technology since 2009, when the market paid little attention to this complex technology with limited initial demand. From the first generation HBM to HBM3E, this high-stakes gamble has lasted nearly 13 years, only reaching a crowning moment with the emergence of ChatGPT.

Image source: AI generated

SK Hynix's journey to today would not be possible without a key external aid. After the dot-com bubble burst in 2001, Hynix was plunged into a debt crisis, with its stock price falling to junk status, even negotiating with Micron Technology to sell, which ultimately failed. For the next decade, the company was under long-term creditor control.

In 2012, SK Group Chairman Chey Tae-won overruled board opposition and acquired a stake through its investment holding subsidiary SK Square for about $3 billion, rebranding it as SK Hynix, and injected large-scale R&D funds. This investment allowed the company to continue advancing HBM technology, which was still a niche market at the time. Currently, SK Square owns about 20% of SK Hynix, making it the largest single shareholder.

Interestingly, SK Square itself has also attempted to enter the crypto market. In 2021, it acquired a 35% stake in Korean crypto exchange Korbit for about 90 billion won and planned to issue its own token, SK Coin. Public reports indicate that after the Terra/LUNA collapse in 2022, market enthusiasm cooled sharply, and SK Coin's issuance plans were shelved, with no substantial progress since.

According to Reuters, citing sources, SK Hynix plans to go public on Nasdaq as early as August this year, which would lower trading barriers for US institutions and passive funds, potentially further attracting capital inflows. NVIDIA CEO Jensen Huang recently stated that future cooperation between NVIDIA and SK Hynix could bring hundreds of billions of dollars in business opportunities to Korea.

Why is capital willing to pay? The mirror of Crypto AI

In this AI wave, the market prefers to pay premiums for segments with actual orders and visible supply bottlenecks. Assets directly involved in AI supply—computing power, memory, electricity—are prioritized because their revenues are quantifiable and barriers verifiable.

HBM capacity is highly concentrated among SK Hynix, Samsung, and Micron, with expansion cycles lasting 2 to 3 years. This physical scarcity is not built on narratives but is locked in by capacity cycles and technological barriers. The valuation logic of the storage industry is also shifting from "cyclical stocks" to "growth stocks."

SK Hynix surpassing Bitcoin in market value is a public statement from the capital market about two kinds of scarcity. The physical barriers are so high that the situation of Crypto AI warrants a reassessment.

The Crypto AI track has been telling a story for the past two years: decentralized computing power will reshape AI infrastructure, and open networks will surpass closed enterprise data centers. The potential of this direction is real, but in light of SK Hynix's current market value, several realities deserve direct attention.

A joint report by 13 universities including Cornell University, IC3, states that the integration of Crypto and AI is still in its early stages, and the hype around this intersection has overshadowed actual progress. Decentralized computing, data markets, and governance mostly remain conceptual.

At the project level, taking Bittensor as the most representative project in the Crypto AI track, its token TAO has fallen 20% over the past three months. Co-founder const posted on X that the project's economic incentives are still dominated by the core team, who choose to maintain centralization for rapid iteration, and expect it will take another year and a half to complete core mechanism development. In other words, their underlying mechanisms are still being patched.

Closer to hardware, crypto mining companies are also in a tough spot. According to Galaxy Research data, Bitcoin miners are entering a "surrender phase," with current network difficulty down over 20% from its all-time high, marking the largest retracement since China's crackdown on Bitcoin mining in 2021, with some miners continuing to exit the network or shut down equipment.

In pursuit of transformation, miners like Core Scientific, TeraWulf, Hut 8 have announced moves into AI and high-performance computing. But according to VanEck, this transition faces a short-term funding gap of about $50 billion, with long-term capital needs around $221 billion, and currently only about 25% of leased AI capacity has been delivered—companies missing construction milestones are facing downgrades from investors.

The IC3 report from Cornell and others also notes that the integration of Crypto and AI remains in its early stages, with most hype still in the conceptual phase.

On the funding front, Arthur Hayes recently published an article titled "Reality Test," pointing out that since ChatGPT's release in 2022, the AI industry has issued about $1.5 trillion in debt, roughly matching the increase in US dollar M2 during the same period—AI has almost absorbed all new liquidity, and Bitcoin has never had such an opportunity. Hayes believes this isn't just about "AI losing funds and flowing back into crypto." The upcoming IPOs of Anthropic and OpenAI will further siphon market funds, and once the AI bubble bursts, bank credit tightening will also restrict liquidity, leading to Bitcoin being sold off along with AI.

Since the second half of last year, many traders active in crypto markets have shifted their focus to US and Korean stocks, chasing AI hardware trends. The logic behind capital flowing into AI infrastructure is straightforward: real orders, physical barriers, quantifiable profit margins.

This certainty is the fundamental reason why current capital is willing to pay high premiums, and the AI narrative in crypto markets lacks this certainty.

In other words, the current benefits of AI infrastructure are more likely to be captured by entities with technological barriers and real supply capacity. Crypto networks need to better define their position within the value chain during this process.