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.
Ethereum's L1 scaling redefines L2s from pure throughput solutions to specialized platforms, while AI agents introduce a new, autonomous layer of on-chain activity.
Investigate L2s that offer unique features or cater to specific enterprise needs beyond just low fees.
The future of crypto involves a more performant Ethereum L1, specialized L2s, and a burgeoning agentic economy.
The rapid rise of autonomous AI agents demands a decentralized trust layer. Blockchains, initially an "internet of money," are now becoming the foundational "internet of trusted agent commerce," providing verifiable identity and reputation essential for multi-agent economies. This shift moves beyond simple payments to establishing a credible, censorship-resistant framework for AI-driven interactions.
Integrate ERC-8004 into agent development. Builders should register their AI agents on ERC-8004 to establish verifiable on-chain identity and reputation, attracting trusted interactions and avoiding future centralized platform fees or censorship.
The future of AI commerce hinges on decentralized trust. ERC-8004 is the foundational primitive for this, ensuring that as AI agents become more sophisticated and transact more value, the underlying infrastructure remains open, fair, and resistant to single points of control. This is a critical piece of the puzzle for anyone building or investing in the agent economy over the next 6-12 months.
Agentic AI is not just a tool; it's a new layer of abstraction for decentralized networks. It shifts the barrier to entry from deep technical and crypto-specific knowledge to strategic prompting and resource allocation, accelerating network participation and value accrual.
Experiment now. Deploy a hosted agentic AI like OpenClaw (via seafloor.bot) with a small budget to understand its capabilities in a controlled environment. Focus on automating complex setup tasks within decentralized AI protocols like Bittensor to gain firsthand experience before others.
The rise of agentic AI agents will fundamentally reshape how individuals and organizations interact with and profit from decentralized AI. Those who master agent orchestration and "skill" development will capture disproportionate value as these systems become the primary interface for programmable intelligence and capital.
AI's gravitational pull on talent and capital is forcing crypto to mature beyond speculative tokenomics, transitioning focus from "meme value" to demonstrable product-market fit and real-world utility.
Identify and invest in projects building at the intersection of crypto and AI, or those creating "net new" applications that abstract away crypto complexity for mainstream users, especially in areas like identity or fintech.
This bear market is a necessary, albeit painful, reset. It's a time for builders to focus on creating tangible value and for investors to seek out projects with genuine utility, as the era of easy speculative gains is over.
The commodification of AI compute, driven by decentralized networks, is shifting power from centralized data centers to globally distributed, incentive-aligned miners. This creates a more efficient, resilient, and cost-effective foundation for intelligence.
Explore building AI agents and applications on Shoots' expanding platform, leveraging their TEEs and end-to-end encryption for privacy-sensitive use cases. The "Sign in with Shoots" OAuth system offers a compelling way to integrate AI capabilities without upfront compute costs.
Shoots is not just an inference provider; it's building the foundational infrastructure for a truly decentralized, private, and intelligent internet. Over the next 6-12 months, expect to see a proliferation of sophisticated AI agents and applications built on Shoots, driven by its unique blend of incentives, security, and global compute.
The Macro Shift: Ethereum pivots from a "rollup-centric" vision to a multi-faceted approach: a powerful, ZKVM-scaled L1 coexists with a diverse "alliance" of specialized L2s. This adapts to technical realities and renews L1's core focus.
The Tactical Edge: Builders should prioritize differentiated L2 solutions or contribute to L1's ZKVM scaling. Investors should evaluate L2s based on distinct utility and symbiotic relationship with Ethereum.
The Bottom Line: Ethereum's market leadership remains, but this pivot signals a pragmatic roadmap. The next 6-12 months will see rallying around L1 ZKVM scaling and clearer L2 roles, demanding sharper focus on where value accrual and innovation occur.
