The memory aspect of semiconductors today has gotten so extreme. Stuff is so expensive that people are simply not able to make lower-end equipment or like devices anymore. And this is like killing everything, right?
AI chips make like 65% operating margins and gaming does like 40%. So obviously from a business perspective it doesn't really make sense to put too much effort into GPUs which is kind of sad you know because what happened to the rest of us you know everything is like AI.
Meta's platform of apps has 3.5 billion daily active users, and they make something like I think it's like $200 a year off of each user in advertising, which just goes to show that like for every person in the world, there's a lot of companies that want to sell them something.
The AI era is fundamentally reorienting the semiconductor industry from consumer-driven volume to enterprise-driven performance and specialized memory. This means sustained, massive capital expenditure from hyperscalers will continue to be the primary growth engine.
Invest in companies providing specialized memory (HBM, high-density NAND) and custom silicon solutions for AI workloads. These components are the bottlenecks and profit centers for hyperscalers.
The AI infrastructure buildout is far from over. Expect continued, accelerating investment in compute and memory through 2027 and beyond, creating a "rising tide" for the entire semiconductor supply chain.
AI's insatiable demand for compute and memory is fundamentally re-prioritizing semiconductor manufacturing, shifting capacity and R&D from consumer products to high-margin data center components. This creates a new economic reality where memory is the bottleneck and a strategic asset.
Invest in companies positioned to supply high-performance memory (HBM, advanced DRAM, NAND) or those hyperscalers with clear, high-margin internal monetization paths for their AI capex (e.g., advertising-driven models).
The AI infrastructure buildout is far from over, with hyperscalers projecting continued, accelerating capex into 2027 and beyond. This sustained investment will keep memory prices elevated and drive innovation in optical interconnects and custom silicon, creating both challenges for consumers and immense opportunities for strategic investors and builders.
AI's pervasive influence is fundamentally re-architecting the semiconductor supply chain, shifting investment from consumer-grade components to high-margin, specialized AI memory and compute, creating a sustained demand cycle.
Invest in companies positioned to capitalize on the broad memory demand, from HBM manufacturers to NAND suppliers, and those hyperscalers with clear, high-margin monetization paths for their AI infrastructure.
The AI infrastructure buildout is far from over, with hyperscalers committing hundreds of billions annually. This sustained investment will continue to drive semiconductor prices and innovation, making memory and specialized compute the critical bottlenecks and opportunities for the next 3-5 years.
Skyrocketing Costs: GDDR7 prices have quadrupled in the last year, with DRAM contract prices doubling in a single quarter. This means the memory (VRAM) now accounts for 80% of a gaming GPU's bill of materials, making consumer GPU manufacturing increasingly unprofitable.
AI's Profitability: AI chips offer significantly higher operating margins (65%) compared to gaming GPUs (40%). This incentivizes companies like NVIDIA to focus on data center AI, meaning less investment in consumer products and a clear business rationale for the current market dynamics.
Enterprise Skepticism: Wall Street is wary of Microsoft's AI capex due to longer enterprise sales cycles and less immediate ROI compared to advertising-driven models. This suggests investors are prioritizing quick, high-margin returns in the current AI gold rush.
The memory aspect of semiconductors today has gotten so extreme. Stuff is so expensive that people are simply not able to make lower-end equipment or like devices anymore. And this is like killing everything, right?
Capex Surge: Google, Meta, Amazon, and Microsoft are collectively committing over $600 billion in capex for 2026, a 70% average increase. This massive investment is primarily directed at building out AI data centers, compute, memory, and networking infrastructure.
NAND's Moment: Nvidia's Vera Rubin platform will feature over 1,152 terabytes of NAND per rack, with Morgan Stanley estimating Reuben alone will consume 13% of global NAND supply by 2027. This highlights the critical role of massive, cheaper storage for context memory and KV cache in scaling AI.
The memory aspect of semiconductors today has gotten so extreme. Stuff is so expensive that people are simply not able to make lower-end equipment or like devices anymore. And this is like killing everything, right?
We're in an era of finding a use case for something that just requires so much memory. This I I don't see it changing in the immediate future.
AI chips make like 65% operating margins and gaming does like 40%.
AI's integration into core business models is driving hyperscalers to commit unprecedented capital to infrastructure, shifting semiconductor demand from consumer-driven cycles to enterprise-grade, high-margin AI components.
Investigate memory manufacturers and specialized AI silicon providers, as their products are becoming the foundational bottleneck and highest-margin components in the AI infrastructure buildout.
The AI capex spend, projected to exceed $600 billion in upcoming years, is a rising tide lifting all semiconductor boats. Understanding where this capital flows—from HBM to NAND and custom silicon—is crucial for positioning your portfolio and product roadmap for the next half-decade.
AI's computational hunger is fundamentally re-architecting the semiconductor industry, shifting focus from consumer-driven volume to high-margin, specialized memory and compute for hyperscalers. This means a sustained, elevated demand for advanced silicon, with traditional consumer markets becoming a secondary concern.
Invest in companies providing core AI infrastructure components—HBM, advanced NAND, and custom silicon design capabilities—or those hyperscalers with clear, high-margin monetization paths for AI, like advertising.
The AI infrastructure buildout is far from over, with hyperscalers projecting continued, accelerating capex into 2027 and beyond. This sustained investment will keep memory prices high and demand for specialized AI hardware robust, creating a new economic reality for tech investors and builders.
Bitcoin's Bull Run is Just Starting: Driven by broad adoption and macro uncertainty, Bitcoin has hit "escape velocity" with significant upside potential.
Regulatory Winds Have Shifted: The impending Genius Act and a more crypto-friendly SEC are set to unleash a wave of innovation and institutional participation.
Tokenization & AI are Converging: The tokenization of real-world assets, especially equities, and the build-out of AI infrastructure (often by crypto-related entities) are major growth vectors.
**Infrastructure is the New Frontier:** Prioritize crypto ventures using blockchain as a foundational layer to innovate and compete with Web2, moving beyond purely crypto-centric applications.
**Solve Real Problems, Not Chase Hypotheses:** True PMF stems from addressing tangible user pain points; market creation is often a byproduct of successful problem-solving, not an initial goal.
**Large Markets Fuel Pivots:** While a sharp focus is vital, building within a substantial market provides the necessary runway and adjacent opportunities critical for navigating the path to PMF.
UX is King: Seamless, integrated user experiences (like Hyperliquid's or a desired "Robin Hood for crypto") will win, as fragmentation (EVM L2s) breeds user frustration and churn.
Solana's Ascent: Alpenlow’s 150ms finality and zero voting costs significantly enhance Solana's competitive edge, driven by an "underdog" culture of relentless improvement.
ETH's Identity Search: Ethereum needs decisive leadership and a unified technical/narrative strategy to counter fragmentation and challengers; price pressure often serves as its main catalyst for action.
**Hyperliquid (Hype) is King:** Flood states, "It's the only asset that matters in crypto other than Bitcoin... Nothing else makes money," citing its strong fundamentals and mispricing.
**L1s are Uninvestable Commodities:** Focus on applications and frontends that directly serve users; L1s are a race to the bottom on fees and vulnerable to tech disruption.
**Builder Codes Fuel an Ecosystem:** Hyperliquid's permissionless monetization will attract a wave of development, creating a moat through network effects and specialized user experiences.
Treasury Tactics: The "treasury company" model is the new "low float, high FDV" game, but relies on continued premium valuations and favorable debt markets; watch out for stress when debt matures.
Sui's Pragmatism: Sui’s handling of the Cetus hack signals that newer chains may prioritize decisive action and recovery over decentralization purity in crises, a trend likely to continue.
Solana's Evolution: Solana’s major consensus upgrade, developed by former critics, showcases a pragmatic, engineering-first approach focused on performance and validator accessibility, potentially strengthening its L1 position.
Crypto Delivers Utility: Stablecoins move trillions monthly, proving crypto's real-world value beyond speculation for fast, cheap global payments.
AI Rewrites Web Economics: AI's direct-answer capability breaks the old ad-traffic model. Crypto offers tools to build the new economic "covenant" required.
Bet on Category Kings: Tech markets are "winner-take-all." Focus on the dominant player in any credible category, especially those led by founders with unique, "earned secrets."
