Nvidia CEO Warns Computing Demand "Skyrocketing" as RAM Prices Surge, Squeezing Crypto Miners - Brave New Coin

AI Computing Demand Reaches Critical Levels

Speaking at Nvidia’s CES 2026 keynote on January 5, Huang emphasized that AI computational requirements are growing exponentially. “The amount of computation necessary for AI is skyrocketing,” Huang stated, noting that AI model parameters increase by a factor of 10 every single year while computing demands continue accelerating.

Nvidia confirmed its next-generation Rubin platform is in full production and remains on schedule. The platform combines six chips including the Rubin GPU and Vera CPU, designed to work together, delivering up to five times greater AI inference performance compared to previous models. The company estimates between $3 trillion and $4 trillion will be spent on AI infrastructure over the next five years. Rubin-based products will be available from partners in the second half of 2026.

Earlier in October 2025, Huang told CNBC that computing demand had risen “substantially” in the prior six months as AI models evolved from simple question-answering to complex reasoning. This development necessitates massive GPU deployments across data centers, with Nvidia’s latest systems consuming up to 1,400 watts per GPU and requiring sophisticated liquid cooling systems.

The surge in AI computing has created what Huang described as “two exponentials happening at the same time” – rapidly advancing AI capabilities alongside mushrooming demand for processing power. This dual expansion is reshaping global semiconductor manufacturing priorities and creating ripple effects across multiple technology sectors.

Historic RAM Price Crisis Grips Tech Industry

The AI boom has triggered an unprecedented memory shortage, with RAM prices experiencing their steepest increases in industry history. Kingston Datacenter SSD Business Manager Cameron Crandall reported a 246% increase in NAND wafer prices compared to Q1 2025, with 70% of that spike occurring in just 60 days.

Memory prices for DDR5 RAM modules have more than doubled in some cases, with consumer products that previously cost a few hundred dollars now approaching four-digit price tags. Team Group’s General Manager Gerry Chen stated that December 2025 contract prices for DRAM increased 80% to 100%, calling this the start of a “multiyear memory upcycle.”

The shortage stems from manufacturers reallocating production capacity from consumer DRAM toward high-bandwidth memory (HBM) used in AI accelerators. Each gigabyte of HBM consumes roughly three times the wafer capacity of DDR5, according to industry analyses. AI workloads are projected to consume nearly 20% of global DRAM supply by 2026.

Micron Technology announced it will exit the consumer Crucial brand by early 2026, citing the need to prioritize “larger, strategic customers in faster-growing segments.” This decision leaves Samsung and SK hynix as the only major suppliers of consumer DRAM. Industry experts warn prices will continue rising throughout 2026, with new fab capacity not expected until 2027 at the earliest.

The memory crisis has created panic buying among consumers and businesses attempting to secure inventory before further price increases. Retailers in Japan’s Akihabara electronics district instituted purchase limits on RAM and SSDs in late October 2025 to prevent hoarding. Some retailers offered “memory certificate” deals allowing customers to pay deposits to reserve RAM at 2025 prices for 2026 delivery.

Crypto Mining Faces Intensifying Hardware Competition

Bitcoin miners find themselves caught in an escalating battle for computing resources as AI companies secure long-term contracts with chip manufacturers. The competition has created what analysts describe as the “harshest margin environment of all time” for cryptocurrency mining operations.

Bitcoin network hashrate recently reached 1.032 zettahashes per second, with mining difficulty at 148.26 trillion heading into 2026. Despite increased computational competition, mining profitability has declined sharply. JPMorgan reported that Bitcoin network hashrate fell for the second consecutive month in December 2025, dropping 3% to 1,045 exahashes per second, while daily block reward revenue per exahash hit record lows, down 32% year-over-year.

Nvidia’s dominance in GPU supply for both AI and crypto mining workloads means its hardware allocation decisions directly impact mining operations. The company has prioritized AI data center contracts over other applications, with multi-year agreements reportedly accounting for the majority of SK hynix’s HBM output through 2026.

The introduction of the Rubin platform represents a structural upgrade in AI compute density but also signals a shift in Nvidia’s hardware optimization priorities. The platform’s chips increasingly target large-scale AI training and inference workloads rather than general-purpose computing applications, limiting their utility for cryptocurrency mining operations.

Bitcoin Miners Pivot to AI Infrastructure

Several major Bitcoin mining companies are transforming their business models to capitalize on AI infrastructure demand. IREN Limited secured a $9.7 billion contract with Microsoft for AI computing power in November 2025, sending its stock soaring 25%. The company plans to deploy Nvidia GB300 processors at its 750-megawatt Texas facility.

JPMorgan identified a crucial nine-month window for Bitcoin miners to secure contracts with U.S.-based hyperscalers and AI startups. The report noted that equipping a 100-megawatt site with advanced GPUs requires approximately $3 billion, highlighting the massive capital requirements for miners attempting to diversify into AI services.

Cipher Mining signed a $3 billion deal with Google-backed Fluidstack in September 2025, with Google securing a 5.4% equity stake. TeraWulf, Hut 8, and Core Scientific have all announced similar pivots toward AI hosting services, leveraging their existing power infrastructure and data center expertise.

The transformation represents a strategic shift for mining companies that previously focused exclusively on cryptocurrency. AI infrastructure contracts offer more predictable revenue streams compared to volatile Bitcoin mining returns, which fluctuate based on cryptocurrency prices and network difficulty adjustments.

Bitcoin mining companies collectively control access to more than 14 gigawatts of power across North America, according to Bernstein analysts. This existing infrastructure positions them as valuable partners for tech giants facing power shortages and long permitting timelines for new data center construction.

Market Implications and Future Outlook

The convergence of AI demand and memory shortages is fundamentally reshaping both industries. Technology companies including Google, Amazon, Microsoft, and Meta have placed open-ended orders with memory suppliers, indicating they will accept available supply regardless of cost. This corporate buying power further disadvantages smaller players, including individual miners and mid-sized operations lacking capital to compete for scarce resources.

Hardware scarcity and rising prices are forcing Bitcoin miners to make critical strategic decisions. Mining companies without access to cheap power or AI partnerships face mounting cost pressures as profit margins compress. The April 2024 Bitcoin halving already reduced block rewards from 6.25 to 3.125 Bitcoin, cutting miners’ primary revenue source in half.

Industry analysts project memory prices will remain elevated throughout 2026, with relief dependent on new fab capacity coming online. Micron’s $9.6 billion Hiroshima HBM facility construction is expected to begin around May 2026, with first output anticipated in 2028. Samsung is accelerating its Pyeongtaek expansion with billions in investments, but company executives have indicated HBM and high-margin enterprise DRAM will receive priority through 2027.

For cryptocurrency miners, the immediate outlook involves difficult choices between maintaining mining operations at reduced profitability, investing heavily in AI infrastructure pivots, or exiting the industry entirely. CoinShares analysts noted that building and operating a Bitcoin mine typically costs around $700,000 to $1 million per megawatt, whereas an AI data center can cost up to $20 million per megawatt, reflecting the higher redundancy and reliability requirements for AI workloads.

The shift from commodity DRAM to specialized HBM production represents a potentially permanent reallocation of global silicon wafer capacity rather than a temporary cyclical shortage. For decades, DRAM and NAND production for smartphones and PCs drove manufacturing priorities. Today, that dynamic has inverted, with AI infrastructure becoming the primary driver of semiconductor investment and capacity allocation.

The Silicon Crossroads

The memory supercycle driven by AI demand represents a fundamental transformation in how semiconductor manufacturing capacity is allocated globally. Nvidia’s announcement of skyrocketing computing requirements, combined with RAM prices surging over 200%, creates a new competitive landscape where traditional cryptocurrency mining must either adapt to AI infrastructure hosting or face increasingly difficult economics. The next 18 months will determine which mining operations successfully navigate this transition and which become casualties of the AI revolution’s insatiable appetite for computing resources.