Dwarkesh Podcast

Investors should prioritize exposure to the Data Preparation and RLHF (Reinforcement Learning from Human Feedback) sectors, as companies like Surge AI and Mercor are essential "picks and shovels" for AI labs. While open-source models are rapidly closing the gap with frontier models, the primary investment moat remains proprietary, expert-level human datasets rather than just software architecture. In the autonomous vehicle and robotics space, Tesla and Waymo are the high-conviction plays as they use massive data "brute force" to overcome current learning efficiency gaps. Despite automation fears, demand for human Software Engineers is projected to increase through 2027, suggesting investors should favor firms that use AI to augment professional productivity rather than replace it. For those tracking fintech, Mercury is a leading private play in AI-native banking through its automated financial management tools.

Avoid investing in Emerging Markets or jurisdictions that have recently experienced a government overthrow, as historical "regime change" dynamics suggest a high probability of multi-decade volatility. Prioritize sovereign debt and equities in nations with Institutional Legitimacy, using the age of a country’s constitution as a primary proxy for long-term capital protection. Seek out companies with high Soft Power and "cultural capital," as these intangible assets act as a defensive moat and low-cost diplomatic tool during periods of global conflict. In markets with weak legal systems, evaluate the strength of a firm’s Patronage Networks and political alliances rather than the written law, as these connections often dictate actual business outcomes. Maintain high Liquidity and diversification to hedge against "Fortune" or black swan events, acknowledging that even the most perfect strategic plans only control roughly 50% of the eventual outcome.

Investors should prioritize the S&P 500 (SPY) as the primary vehicle to capture broad productivity gains as AI integrates into every sector of the economy. To capitalize on the immediate scarcity of processing power, maintain exposure to AI hardware and infrastructure leaders like NVIDIA (NVDA) and the broader Compute Index. Shift long-term portfolios toward the "relational sector," focusing on high-touch industries like healthcare, luxury hospitality, and specialized professional services where human empathy commands a premium. Avoid "commodity" white-collar roles vulnerable to automation, instead favoring senior management and roles requiring high-stakes accountability. Monitor political stability and labor share data closely, as any significant rise in unemployment could trigger sudden regulatory shifts or changes in tax law.

Investors should maintain high conviction in NVIDIA (NVDA) as their new B100/B200 chips achieve a 3x performance boost in FP4 precision, offering exponential efficiency gains over competitors. For exposure to specialized AI training at scale, Alphabet (GOOGL) remains a top pick as their TPU architecture minimizes "data movement taxes" more effectively than general-purpose hardware. Keep a close watch on the private markets for Maddox, a startup developing a "splittable systolic array" that could bridge the gap between NVIDIA’s flexibility and Google’s raw efficiency. A critical metric for evaluating any semiconductor investment is the ratio of compute area to data movement area, as hardware that minimizes overhead will lead in performance-per-watt. The industry-wide shift toward FP4 precision is the most time-sensitive trend, favoring companies that can maintain accuracy while utilizing the quadratic physical area savings of lower-bit widths.

Investors should maintain long-term exposure to Alphabet (GOOGL) as they pivot from consumer AI to solving "intractable" high-value problems in biology and physics via AlphaFold and AlphaTensor. NVIDIA (NVDA) remains a high-conviction play due to "inference scaling," a trend where AI models require massive sustained compute power to "think" during the reasoning phase, not just during initial training. Look for opportunities in Automated AI Research software that provides verification for autonomous scientific discoveries, as AI begins to replace junior research engineers in hyperparameter optimization. The rapid commoditization of AI training—where frontier capabilities that once cost millions now cost under $10,000—suggests a shift in value toward companies with proprietary data quality rather than just raw compute. In the robotics sector, watch for firms utilizing Foundation Models and "Dagger" algorithms to amortize complex physical movements into efficient, real-time neural network passes.

The "industrialization" of genetic sequencing is shifting value toward "picks and shovels" providers like Illumina (ILMN) and Pacific Biosciences (PACB), which facilitate high-throughput data generation. Investors should prioritize firms specializing in Targeted Enrichment and Library Preparation, as specialized chemistry is now more critical than raw sequencing power for extracting high-quality data. The transition of genomics into a "Big Data" field makes Cloud Compute and Bioinformatics infrastructure essential, creating opportunities in firms that provide specialized ML environments for life sciences. Precision Medicine companies utilizing Polygenic Risk Scores (PRS) are well-positioned to capitalize on new data identifying thousands of genetic markers for chronic diseases like Type 2 Diabetes. Long-term growth is expected in Nutrigenomics and CRISPR-based AgTech, focusing on aligning modern diets and livestock with human evolutionary biology.

Investors should maintain a high-conviction position in NVIDIA (NVDA), specifically focusing on the transition to the Blackwell NVL72 and upcoming Rubin architectures which solve critical "scale-up" networking bottlenecks. Beyond raw chips, look for opportunities in the "cabling and switching" sector, as the physical density of interconnects and NVLink technology is now the primary constraint on AI model scaling. A significant portion of hyperscaler CapEx is being consumed by High Bandwidth Memory (HBM), making companies that specialize in CXL (Compute Express Link) and tiered memory management essential for reducing costs. Efficiency-first architectures like DeepSeek demonstrate that "sparse" models (Mixture of Experts) will dominate the market by offering frontier-level performance with significantly higher profit margins. Finally, monitor the shift from training-heavy to inference-heavy hardware, as bespoke chip startups focusing on memory bandwidth rather than just raw math speed are poised to capture the next wave of AI infrastructure spending.

NVIDIA (NVDA) remains the top conviction play as it shifts to an aggressive one-year product cycle, leveraging a $100B-$250B supply chain moat that makes it nearly impossible for competitors to catch up. Investors should look toward specialized software providers like Synopsys (SNPS) and Cadence (CDNS), which are poised for a volume explosion as AI agents begin using these tools 24/7. High-bandwidth memory remains a critical bottleneck, positioning Micron (MU) as a primary beneficiary of NVIDIA's massive downstream demand. In the networking and scaling space, Lumentum (LITE) and Coherent (COHR) are key strategic partners to watch as silicon photonics becomes essential for future AI infrastructure. Finally, the ultimate constraint on this growth is power; therefore, any long-term AI portfolio must account for the energy sector and electrical infrastructure required to fuel "AI Factories."

Prioritize investments in companies with massive, proprietary experimental datasets, as AI breakthroughs in physical sciences are 90% dependent on high-quality data moats like the Protein Data Bank. Focus on Software Engineering tools that assist in high-level system design and architecture, as LLMs are rapidly commoditizing basic code syntax. Treat Synthetic Biology and Biomimicry as high-conviction plays, as these sectors are effectively "translating" nature’s complex biological machines into scalable engineering assets. View Quantum Computing as a long-horizon "deep tech" investment, focusing on firms developing new algorithms beyond simple encryption-breaking. Use the vibrancy of Open Source communities and Preprint activity on platforms like arXiv as leading indicators to identify the next commercial breakthroughs before they hit the mainstream market.

As AI drives the cost of idea generation toward zero, the highest conviction investment theme is in formal verification and evaluation layers that can validate AI outputs. Focus on companies integrating with Lean or building systems that bridge the gap between natural language AI and formal logic to eliminate "hallucinations" in mission-critical fields. Investors should prioritize firms with proprietary, high-precision "clean" datasets, as these serve as defensive moats against generic models. Look for Quant Hedge Funds and data extraction specialists that employ physicists to find signals in the current "deductive overhang" of unanalyzed big data. Finally, maintain exposure to productivity tools like LabelBox or Wolfram Alpha that automate research drudgery, while remaining cautious of over-optimized companies that have eliminated the "human slack" necessary for breakthrough innovation.

NVIDIA (NVDA) remains a top-tier conviction play as it secures 70% of TSMC’s 3nm capacity through 2027, with the transition to Blackwell and Rubin architectures expected to drive massive margin expansion. Investors should view ASML (ASML) as the ultimate industry bottleneck; its monopoly on EUV lithography tools makes its shipping guidance the primary leading indicator for the entire AI sector's growth ceiling. A severe "memory crunch" is expected through 2027, making SK Hynix, Samsung, and Micron (MU) high-conviction buys as AI demand cannibalizes standard chip supply and drives up prices. To play the critical energy constraints facing data centers, look toward infrastructure providers like GE Vernova (GEV), Vertiv (VRT), and Eaton (ETN) that specialize in modular power and cooling solutions. Finally, monitor the massive $600 billion combined CapEx from Microsoft (MSFT), Google (GOOGL), Amazon (AMZN), and Meta (META), as those who locked in long-term compute contracts early now hold a significant margin advantage over latecomers.

The U.S. government’s designation of Anthropic as a supply chain risk suggests private AI labs may face significant valuation headwinds if they refuse to comply with military surveillance demands. Investors should favor "bridge" companies like Palantir (PLTR) that are already deeply integrated with the Department of Defense and can navigate the friction between ethical AI and military utility. NVIDIA (NVDA) remains the essential "neutral arms dealer" in this conflict, though it faces increasing federal pressure regarding chip allocation and customer vetting. The rapid 10x annual cost deflation in AI inference makes multimodal AI software and Open Source models (like Meta’s Llama) the primary drivers for the next wave of mass surveillance infrastructure. Finally, prioritize companies with pre-existing, permitted data centers and independent power agreements, as the government is increasingly using energy permitting as a "soft power" lever to control non-compliant AI firms.

To capitalize on the current Information Revolution, investors should shift focus from flashy end-products to the "plumbing" and distribution hubs, such as AI data centers and logistics infrastructure, that allow new technologies to scale. Prioritize companies like NVIDIA (NVDA) that foster open ecosystems and industry standards, as collaborative platforms historically capture more market value than "black box" or siloed competitors. Look for "deflationary" plays where expensive inputs are being replaced by cheap commodities, specifically targeting firms innovating in battery chemistry to reduce reliance on rare earth metals. High-conviction capital should be allocated to jurisdictions with strong institutional checks and balances, as political "friction" acts as a long-term safeguard for property rights against arbitrary leadership. View Artificial Intelligence not as a standalone bubble, but as a multi-decade compounding phase of the digital age that rewards companies controlling the speed and flow of information.

With a 50% chance of a major AI breakthrough in the next 1-3 years, trillions of dollars are being invested into the underlying infrastructure. The most direct way to invest in this trend is through the "picks and shovels," particularly chipmakers who supply the essential hardware for the AI build-out. Consider large cloud providers like Microsoft (MSFT) and Google (GOOGL), as they are the biggest buyers of AI chips and leaders in developing frontier AI models. The exponential growth in data centers also creates a significant opportunity for companies involved in the power grid and energy generation. For a longer-term strategy, look for companies in Biotech and Software that are effectively integrating AI to revolutionize their industries.

Consider investing in memory chip makers like Micron (MU), as memory is identified as the biggest bottleneck for scaling AI with prices reportedly "going ballistic." The entire semiconductor sector, including foundries like TSMC (TSM), is also experiencing demand that far outstrips supply, creating a bullish outlook. A major theme for the next few years will be the electricity shortage needed to power AI data centers, creating a systemic risk for hardware growth. This power constraint creates a long-term opportunity in the energy sector, particularly within the gas turbine supply chain, which is reportedly backlogged through 2030. While demand for NVIDIA (NVDA) chips is strong, be aware that this electricity bottleneck could create a growth ceiling for the company towards the end of this year.