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.
**Memecoins Were a Trojan Horse:** The speculative frenzy was a catalyst that massively accelerated DEX adoption and forced millions of users to finally learn how to use self-custody wallets and on-chain tools.
**Prepare for Thousands of Stablecoins:** Every company with deposits will likely issue its own "branded money." The next major infrastructure battle will be building the interoperability layers—the "Visa for stablecoins"—to manage this fragmented liquidity.
**The Real Stablecoin Opportunity is Global:** The next frontier isn't another USD competitor, but non-USD stablecoins tied to high-yield foreign currencies, which will unlock the creation of on-chain foreign exchange (FX) markets.
DEXs are Eating the World. The on-chain asset explosion has permanently shifted trading gravity. Centralized exchanges must now integrate with DeFi or risk becoming irrelevant islands.
Stablecoins are the New Gift Cards. The move to "branded money" will create a fragmented landscape. The next billion-dollar opportunity is not in issuing another stablecoin, but in building the interoperability rails that make them all work together seamlessly.
Distribution is the New Defensibility. As stablecoin issuance becomes commoditized, the winners will be those with massive distribution networks (like Stripe) who can embed their currency into everyday user flows.
FHE is crypto’s HTTPS moment. Just as HTTPS made secure browsing the default, FHE is positioned to bring end-to-end encryption to all blockchain transactions, solving a fundamental flaw without forcing users to change their behavior.
Privacy is coming for your wallet, not a new chain. The "holy grail" is integrating confidentiality directly into the user's existing workflow on mainnet Ethereum. Forget bridging; the future is an "incognito mode" for your current assets.
Institutional demand will drive retail privacy. The need for financial institutions like JPMorgan to protect their trades on-chain is the catalyst that will finally make robust privacy tools a standard feature for everyone.
**Stop Applying Linear Valuations to Exponential Tech.** Judging Ethereum on its P/E ratio is like criticizing Amazon in 1999 for its lack of profits. It’s a category error. Value chains based on their probability of capturing a piece of a future trillion-dollar system.
**The Prize Is Worth Winning.** The entire investment case for new L1s hinges on the belief that incumbents like Ethereum and Solana are immensely valuable. If they are, then a small probability of becoming the next one justifies a multi-billion dollar valuation today.
**Zoom Out and Believe.** The current market is trapped in short-term cynicism. The real alpha comes from adopting a Silicon Valley mindset over a Wall Street one, recognizing that you are living through a technological revolution on par with the early internet.
Weaponize cringe for distribution. The ‘Choose Rich Nick’ model proves that being the butt of the joke is a powerful growth hack. Manufacturing moments that invite mockery creates a viral loop of outrage and engagement that funnels attention to the core business.
Authenticity is a liability. The most successful stunts are meticulously planned fabrications. From fake girlfriends to staged yacht expulsions, the goal isn't to be real but to create a compelling narrative that the internet can’t ignore.
Success hinges on ambiguity. The content is designed to polarize. Its virality depends on a split audience: one half gets the joke and celebrates the performance, while the other half takes it at face value, fueling the outrage machine that drives impressions.
Fintech is the New On-Ramp. Giants like Klarna are adopting stablecoins for economic utility, not speculation. This signals a new wave of adoption driven by real-world efficiency gains.
Re-evaluate Your Valuations. The massive valuation gap between a fintech like Klarna and an L1 like Solana forces a critical question: will value accrue to the rails or the businesses that use them to serve hundreds of millions of customers?
Distribution is Undefeated. Robinhood’s move to sideline its partner Kalshi proves that owning the customer relationship is the ultimate moat, a crucial lesson for infrastructure projects reliant on third-party distribution.
