Ideas
Nasdaq/S&P beat most risk asset classes
Most venture funds are playing momentum games that are not durable. All risk asset classes have been bad relative to simply buying the Nasdaq or S&P 500 over the past decade; unless an investor can be top decile, passive index exposure is the superior benchmark and alternative.
Bio is Compound's heaviest investment area
Bio is where Compound has invested most heavily over the past three years, with the second fund heavily allocated to bio. He sees opportunities across the sector, including public data-asset companies, left-for-dead small caps, and a reacceleration of experimentation and biomanufacturing.
Most excited about material science
The area he is most excited about is material science: full-stack businesses built around synthesizing novel materials and shipping an end product, with Orbital Materials as the case study.
Wait until AI dust settles
AI is at a maximally obvious and crowded point, so Compound observes and avoids deploying too much private capital now. The better time to invest in AI is likely 2028-2030, after cheap intelligence commoditizes and the post-buildout dust settles.
Orbital Materials ships chemistry-powered cooling
Orbital Materials is Compound's case study and one of its biggest second-fund investments: a foundation model for chemistry, state-of-the-art chemical synthesis, and a material shipped as a full modular data-center cooling product for next-generation GPUs.
Google is structurally advantaged and hated
Everyone hates Google, but he thinks that is wrong. Google has structural advantages across models, cloud/GCP, TPUs, and distribution, ships strong fast and cheap Gemini models, and can compound for a long time. He notes the market treating Google debt as less risky than Treasuries and is explicitly quite bullish.
Hated software offers AI transition dislocations
Technology and software are overly hated because of AI. He looks for software businesses that can transition to usage-based AI models, accepting short-term margin compression for long-term expansion, and management teams that understand AI. This creates dislocations as the market overweights near-term margin pressure.
Public biotech data assets undervalued
A subset of public biotech companies sit on undervalued data assets that can be monetized sooner and at larger scale than analysts appreciate. There is also a long tail of uncovered, left-for-dead small caps that could become valuable as experimentation and biomanufacturing reaccelerate; he avoids single-asset clinical-trial trading.
Short narrative-driven overvalued deep tech
The next decade is about market-cap destruction as narrative-driven flows reward deep-tech companies that sell great stories. Many are overvalued, have near-term catalysts that will reveal weaker growth, or may not exist in two to four years; short them only with a clear catalyst, multiple expression routes, and controlled correlation and size.
Meta best positioned to monetize AI
Meta is probably best positioned to monetize AI regardless of whether it builds frontier models, because its products are deeply embedded and have enormous surface area. It can monetize AI with ease, Reels has executed well, and optional compute expansion could add growth.
Apple is incredibly well positioned
Apple is incredibly well positioned, and the market has increasingly noticed it over the past 16 months. He sees it as another megacap tech beneficiary if tech beta continues.
Neoclouds will disperse on execution
The next generation of neoclouds will show dispersion based on execution competence, financing relationships, and verticalization choices. Moats of competence and execution will matter over the next five years; he likes some and hates others.
AI obvious basket keeps outperforming
He built a 2024 equal-weighted AI obvious basket of about 30 names, including ASML, Micron, hyperscalers, and Alibaba, and has not changed it. It has drastically outperformed, supporting the idea that structurally obvious AI theme baskets can be held as high-beta exposure.
Own crypto tokens that generate fees
He remains bullish on a small subset of crypto projects that can ascribe fees to tokens and compound with tech-like power-law dynamics. This is mostly application-layer and non-money assets in DeFi, DePIN, DeSci, and protocols built on top of L1s.
Still a Bitcoin bull
He remains a Bitcoin bull and still believes the underlying crypto idea of organizing long-term capital around strange or non-obviously large ideas matters. He expects to own interesting crypto assets in the fund over time.
Robotics this time is different
Robotics was an early miss because AI policy learning was too early, but performance scaling has changed: this time is really different. He continues to invest in both vertically specific robots and more generalizable models, with falling cost curves enabling sub-$5k robots and hospital and vertical use cases.
Humanoid deca-trillion narrative is fragile
The deca-trillion humanoid narrative is fragile. Humanoids only make economic sense where they replace the cost of human life, such as military or industrial uses, but industrial is the most penetrated robotics area. The narrative is being used to draw buyers into a basket of companies, and expectations should be more measured.
Alquist wins forward-deployed enterprise robots
Compound is an investor in Alquist, which focuses on forward-deployed enterprise robotics use cases in retail, data centers, and semiconductors. Its platform has many skill sets and can operate around humans in unstructured environments, with a software layer supporting near- and long-term analysis.
Drone defense budget setup worth watching
There are some public drone and drone-defense companies that could become much larger than they are today, and there is a lot of military budget to allocate there. However, he does not have a strong view or investments in drone defense yet.
Drones have a military supercycle
Drone warfare is not fully priced in. Ukraine demonstrates deep strikes causing tens of billions of damage at orders-of-magnitude lower cost, front lines are increasingly unmanned, and drones are only about 1% of roughly $3T annual military budgets. He expects that share to rise and sees a tailwind for the entire drone industry.
This Delphi Digital video, published August 19, 2026,
features Michael Dempsey, José
discussing QQQ, SPY, XBI, Materials science, AI-SECTOR, Orbital Materials, GOOG, IGV, Public biotech data assets, Deep tech public equities, META, AAPL, Neoclouds, BABA, ASML, MU, SKYY, Fee-generating crypto tokens, DEFI, DEPIN, DESCI, BTC, ROBO, Humanoids, Alquist, Drone defense, ITA.
20 trade ideas extracted by AI with direction and confidence scoring.
Speakers:
Michael Dempsey,
José
· Tickers:
QQQ,
SPY,
XBI,
Materials science,
AI-SECTOR,
Orbital Materials,
GOOG,
IGV,
Public biotech data assets,
Deep tech public equities,
META,
AAPL,
Neoclouds,
BABA,
ASML,
MU,
SKYY,
Fee-generating crypto tokens,
DEFI,
DEPIN,
DESCI,
BTC,
ROBO,
Humanoids,
Alquist,
Drone defense,
ITA
