In May 2026, global capital markets witnessed two seemingly parallel yet deeply interconnected price trajectories: On one side, South Korean memory chip giants Samsung Electronics and SK Hynix saw their stock prices reach historic highs, with market capitalizations each surpassing the $1 trillion mark. On the other, decentralized compute and storage tokens—led by Render Network (RENDER) and Filecoin (FIL)—recorded significant gains over the past thirty days.
These two asset groups belong to entirely different market systems: traditional equity markets and crypto asset markets. Their valuation frameworks, investor structures, and liquidity dynamics are fundamentally distinct. Yet, they are being linked by a common thread: a structural shortage of AI computing power.
This is not a coincidental resonance, but a cross-market revaluation triggered by a supply-demand imbalance in the physical world.
The New Storage Giants in the Trillion-Dollar Club
On May 6, 2026, Samsung Electronics’ stock surged nearly 16% intraday, reaching a record 270,000 KRW per share. The stock closed up 14.41% that day, propelling its market cap past $1 trillion and making it the second Asian tech company—after TSMC—to cross this milestone. The rally continued, and on May 27, Samsung’s share price hit another record high, closing up 2.68%, driven by the formal approval of a labor agreement.
Shortly after, SK Hynix’s stock soared 11% on May 27, officially breaking the $1 trillion market cap barrier and becoming Asia’s third member of the trillion-dollar club. Since the start of 2026, SK Hynix’s share price has skyrocketed about 250%. On the same day, the Korea KOSPI Index jumped as much as 5%, triggering a circuit breaker and pausing algorithmic trading for five minutes.
In the US, Micron Technology closed up 19.29% on May 26, with its share price at $895.88 and market cap crossing $1 trillion for the first time—an eightfold increase over the past year. Within just a few weeks, the world’s three leading memory chip makers—Samsung Electronics, SK Hynix, and Micron Technology—each achieved the trillion-dollar milestone.
Meanwhile, in the crypto market, Gate’s market data showed that as of May 28, 2026, RENDER was priced at $2.1023, up 21.68% over the past 30 days; FIL was at $1.0414, up 12.05% in the same period. The price surge of these two decentralized infrastructure tokens closely mirrored the rally in traditional semiconductor stocks.
From GPU Shortage to End-to-End Compute Crunch
The underlying driver behind all this is a sweeping shortage of computing power across the entire AI industry chain.
From late 2025 to early 2026, as large AI models moved from training to large-scale inference deployment, global demand for compute power entered an exponential growth phase. IDC predicts that the number of active AI agents worldwide will grow from 28.6 million in 2025 to 2.216 billion by 2030—an 80-fold increase in five years. According to China’s National Data Bureau, as of March 2026, China’s daily average token calls exceeded 140 trillion, up 1,400 times from 100 billion at the start of 2024.
While demand exploded, supply hit a structural bottleneck.
Goldman Sachs reports that in 2026, global supply gaps for DRAM, NAND flash, and HBM will reach 4.9%, 4.2%, and 5.1% respectively—the highest since 2011. More critically, SEMI China President Feng Li revealed that HBM capacity shortfall is as high as 50% to 60%, and the three major memory manufacturers have decided to allocate 70% of new capacity to HBM and advanced packaging. SK Hynix’s HBM capacity through the end of 2026 has already been pre-committed by major AI customers.
The GPU spot market is similarly strained. One-year H100 contract rental rates rose nearly 40% in six months, from $1.70/hour in October 2025 to $2.35/hour in March 2026. Spot prices climbed to about $4.08/hour. High-end GPUs like NVIDIA’s B200 are almost "invisible" in the domestic market, and H200 spot prices have even inverted, costing more than the newer B200.
This shortage is not a typical cyclical supply-demand fluctuation, but a structural capacity squeeze triggered by the AI boom. The core bottleneck has shifted from "computing too slowly" to "moving too slowly"—GPU performance depends heavily on HBM, and HBM capacity expansion is limited by advanced packaging and silicon via processes. Real large-scale new capacity may not come online until at least 2028.
Three Intersecting Trends Reveal Supply-Demand Dynamics
Overlaying AI compute demand growth, Samsung and SK Hynix stock performance, and RENDER and FIL price changes on a single timeline reveals a clear "demand transmission chain."
Layer One: Underlying demand drivers. In early 2026, Google, Microsoft, Amazon, and Meta collectively announced annual capital expenditures totaling about $725 billion, with a significant portion flowing directly into data center expansion and large model training clusters. Each AI server requires 8–10 times more DRAM and 3 times more NAND Flash than a traditional server. TrendForce data shows that the combined capital expenditure of the world’s nine largest cloud providers is expected to reach $830 billion in 2026, with annual growth rate rising to 79%.
Layer Two: Value capture in semiconductor manufacturing. Samsung Electronics’ Q1 2026 operating profit reached 57.2 trillion KRW, surpassing its full-year 2025 total and up 756% year-over-year. SK Hynix’s Q1 operating profit rose 405% year-over-year to 37.6 trillion KRW, with an operating margin around 72%. HBM4 components are priced at about $700, 20–30% higher than the previous HBM3E generation, with expected operating margins of 50–60%. The three major memory makers now wield unprecedented bargaining power over the world’s largest tech firms, and the memory chip shortage is projected to last through 2027.
Layer Three: Spillover effects in decentralized networks. When centralized compute supply can’t meet demand, capital seeks alternatives. In January 2026, Render Network generated $38 million in monthly revenue, ranking second among global DePIN projects, with 5,600 active GPU nodes. In April 2026, through governance proposal RNP-023, Render Network integrated about 60,000 GPUs from Salad Network, dramatically boosting its capacity for large-scale AI training and rendering. Filecoin’s foundation shifted its 2026 strategy from "expanding supply" to "expanding paid demand," focusing on AI agents, on-chain data, and real-world assets. AI infrastructure demand has nearly booked out all 2026 storage capacity, first driving up traditional hard drive manufacturer stocks, then rotating to their crypto counterparts.
The table below shows the full transmission path of this value flow from upstream to downstream:
| Industry Chain Level | Core Bottleneck | Representative Asset | Market Type |
|---|---|---|---|
| Silicon Wafers & Raw Materials | Advanced process capacity constraints | Samsung Electronics, SK Hynix | Traditional equities |
| Storage & Packaging | HBM capacity gap of 50–60% | Micron Technology | Traditional equities |
| GPU & Compute Hardware | B200 supply tight, delivery cycles extended | NVIDIA | Traditional equities |
| Data Center Hosting | Power & cooling infrastructure shortage | IREN, Cipher Digital | Traditional equities (mining firms transitioning) |
| Decentralized Compute | Aggregating idle GPU resources | Render Network | Crypto market |
| Decentralized Storage | Surging AI data storage demand | Filecoin | Crypto market |
Structurally, decentralized compute networks are not aiming to replace NVIDIA or Samsung, but to provide incremental supply in "edge zones" where the existing supply chain falls short. Their relationship is complementary, not substitutive.
Consensus and Key Divergences
Bullish Narrative
The bullish camp argues that this memory chip supercycle is fundamentally different from any previous cyclical boom. UBS calls it a "once-in-30-years" memory supercycle, raising its 2026 HBM terminal bit demand forecast to 32.9 billion Gb, up 88% year-over-year, and further to 58 billion Gb in 2027. Nomura Securities issued a most optimistic target price of 590,000 KRW for Samsung Electronics, and an even more aggressive 4 million KRW target for SK Hynix.
On the crypto side, bulls believe that persistent GPU shortages provide fertile ground for decentralized physical infrastructure networks. As AI startups face severe GPU supply constraints—because cloud providers prioritize internal teams and top clients—DePIN platforms like Render Network become price setters in the "long tail" compute market. Filecoin also benefits from rising AI data storage costs, accelerating the shift toward decentralized storage.
Bearish Voices
Bearish perspectives deserve attention as well. On Wall Street, skeptics warn that the cyclical nature of the memory chip industry won’t disappear just because of AI. Samsung and SK Hynix are cautious about ramping up DRAM production, fearing that aggressive expansion could lead to oversupply once AI demand stabilizes. The unprecedented concentration risk in Korea’s stock market also raises red flags.
In the crypto market, doubters point out that some DePIN tokens have fallen 94–99% from their historical highs, making them among the worst-performing sectors. The core challenge for DePIN projects is not "insufficient compute," but the negative feedback loop caused by fixed issuance models during demand downturns, leading to compute attrition. Despite RENDER’s strong 30-day performance, it is still down 52.28% over the past year, while FIL dropped 64.03% in the same period, indicating that token prices remain weighed down by overall crypto market volatility.
Industry Impact: Mining Firms’ Transformation and Compute Supply Chain Reshaping
One of the most profound structural impacts of this AI compute shortage is the large-scale transition of Bitcoin mining companies into AI data centers.
By mid-2026, the industry had signed over $70 billion in AI and high-performance computing contracts. Some mining firms expect up to 70% of their revenue to come from AI business by year-end 2026, effectively transforming into data center operators. Hut 8 leveraged Bitcoin as collateral to build a $16.8 billion leasing base. IREN purchased about $1.6 billion worth of NVIDIA Blackwell architecture AI compute systems from Dell Technologies, supporting a previously announced five-year, $3.4 billion cloud compute services contract.
Mining firms transitioning to AI data centers have clear advantages over cloud giants building from scratch: compute infrastructure, power, and sites are already in place, with grid connections, substations, and usable campuses. With some retrofitting or minor upgrades, they can quickly go live and lock in cash flows via multi-year AI contracts.
This trend has triggered two ripple effects in the crypto market: On one hand, mining firms are selling large amounts of Bitcoin to fund AI infrastructure, with publicly listed miners collectively reducing holdings by over 15,000 BTC, putting some pressure on Bitcoin network hash rate and market price. On the other hand, the transformation itself is creating a bridge between the traditional AI compute market and the crypto market, leading to a crossover in valuation logic.
Conclusion
AI computing power is becoming a universal pricing language across asset classes. From Samsung Electronics’ HBM production lines in Pyeongtaek, to the roughly 60,000 idle GPUs aggregated on Render Network, to the AI training datasets hosted on Filecoin, the same demand curve is being priced in different ways.
Traditional equities and crypto markets are not two parallel lines that never meet. When supply-demand imbalances in the physical world reach a large enough scale, value flows bidirectionally along the industry chain—first crystallizing as profit and market cap in semiconductor manufacturing, then spilling over into token price discovery via decentralized networks.
This does not mean RENDER or FIL will replicate the gains of Samsung or SK Hynix, nor does it mean decentralized compute can replace NVIDIA’s GPU clusters. But it highlights a trend: Under the shared theme of AI compute, two previously isolated market systems are building a linkage based on underlying supply-demand logic. Understanding this relationship may be more important than predicting the price trajectory of any single asset.
