Misunderstood Stocks in Data Center Power, Cybersecurity, and Payments | Dean & Deiya Pernas

Episode Summary

Executive Summary: Dean and Daya Pernas argue that AI’s open-weight shift is bearish for frontier-model labs but bullish for hyperscalers, GPU suppliers, energy, bandwidth, and selected software/cybersecurity names. They stress valuation discipline, variant views, and a buy-side research process built on track record, citing several contrarian small/mid-cap picks tied to data-center power, networking, robotics, cyber exposure management, payments, and cross-border money movement.

Main Topics: AI ecosystem and open-weight models: The brothers frame AI as a four-layer stack and argue open-weight models commoditize frontier-model complements, increasing demand for cloud, GPUs, infrastructure, and enterprise AI use rather than destroying the whole ecosystem. Data-center power and energy infrastructure: They see power constraints as the biggest bottleneck in AI infrastructure, highlighting bring-your-own-power solutions, natural gas, microturbines, and portable LNG systems as urgent beneficiaries. Bandwidth and interconnectivity between data centers: They expect data centers to become more distributed, requiring much more bandwidth and fiber-based interconnects across sites, benefiting specialized networking hardware and software providers. Cybersecurity in the age of AI: They argue AI lowers the cost of finding code vulnerabilities, shifting cyber risk from mostly identity/phishing attacks toward traditional exposure and vulnerability management, which favors companies adapting to this new regime. Software dislocation and selective resilience: They reject a blanket bearish view on SaaS, saying AI threatens some software categories but leaves defensible businesses with unique data, integrations, or real-world workflows intact; valuation and SBC matter a lot. Payments and cross-border money movement: They prefer fintechs focused on international transfers and specialized processing over commoditized payment rails, arguing take rates can remain durable where trust, infrastructure, and network effects matter. Investment philosophy and research process: Pernas Research is presented as a buy-side-style research business built around auditable performance, variant perception, frequent idea generation, and portfolio sleeves for core, starter, and speculative positions.

Key Arguments: Open-weight AI models may hurt frontier labs’ pricing power but expand total AI usage, benefiting cloud providers, GPU owners, and other infrastructure layers. Hyperscalers are being forced to spend heavily on AI capex, creating an opportunity for active managers to find beneficiaries while index leaders absorb uncertain ROI. Power is the gating factor for AI buildout; grid limitations and long lead times make behind-the-meter and portable natural-gas solutions attractive. Bandwidth between data centers will need to expand substantially as large monolithic sites give way to distributed facilities that must communicate continuously. AI increases cybersecurity risk by making vulnerability discovery cheaper, so exposure management and asset mapping become more important than pure identity defense. Not all software is doomed: companies with proprietary data, deep integrations, enterprise-grade security/compliance, or real-world workflows can remain durable. Valuation matters throughout the portfolio; the brothers repeatedly prefer cheaper, smaller companies with similar tailwinds to expensive, better-known names. Cross-border fintechs can defend take rates when they own infrastructure, trust, and customer relationships, especially in remittances and business transfers. Stablecoins are viewed as a limited threat because they are still mostly used for crypto activity rather than everyday fiat-like payments. Their edge comes from variant, contrarian views and a willingness to look at hundreds of companies, not from perpetual buy-and-hold investing.

Data Points: Audited portfolio return since inception: Double the S&P 500 - Pernas Research track record since 2017 Gross return since 2017: Over 1,300% - Audited portfolio performance cited at the top of the episode Annualized return since 2017: 32% annualized - Audited performance figure Top 10 index concentration: Close to 30%-40% - Used to argue the index is vulnerable because leaders are expensive and capex-heavy Hyperscaler valuation: Roughly 10x EV/sales on average - Approximate valuation mentioned for major cloud/AI spenders AI share of U.S. energy grid demand by 2030: 20% - Estimated share of U.S. grid demand AI could consume U.S. energy demand baseline: About 4,000 terawatt hours - Historical annual U.S. energy demand level referenced Power-grid queue times: 18 to 24 months - Lead times to get power from the grid for data centers Bandwidth expansion need: About 15x - Estimated increase in bandwidth needs between data centers Capstone market cap: About $300M-$350M fully diluted - Used to frame the microturbine company as a small-cap, speculative holding Capstone capacity utilization: About 10% - Speaker said the company is operating far below potential capacity Capstone peak move: Up 1,131% - From initiation date to peak before the pullback Capstone current gain: Still up 545% - After the sharp decline from highs Stabilis market cap: Around $75M - Used to emphasize speculative sizing Tenable valuation: About 4x EV/sales - Compared with larger cybersecurity peers CrowdStrike / Palo Alto valuation: North of 20x EV/sales - Used to support Tenable as a cheaper cyber exposure-management play Send volume growth for Remitly-like businesses: 30%-40% annually - Illustrated strength in cross-border payments and remittances Revenue growth for cross-border fintechs: About 25% - Referenced alongside send-volume growth Traditional cross-border transfer cost: About 5% - Compared to cheaper fintech solutions Fintech cross-border transfer cost: Closer to 2% - Illustrated cost savings from modern transfer rails PaySafe leverage: $2.5B debt relative to $600M book value - Used to describe balance-sheet risk PaySafe growth profile: High single-digit growth - Described as organically better than it screens Meta spend relative to revenue: Very high / spending heavily for itself - Speaker’s qualitative critique of Meta’s AI capex Meta acquisition of AI talent: Hundreds of millions of dollars - Referenced when discussing poaching from OpenAI/Anthropic

Pivotal Quotes: "We believe this is bullish for enterprises because they can now spend a lot more with AI." — Dean Pernas: Explaining why open-weight AI may increase usage across the ecosystem rather than reduce it "It’s a special time to be an active manager." — Daya Pernas: Describing how concentrated index weights and expensive hyperscalers create stock-picking opportunities "The cost of finding vulnerabilities in code has gone down significantly." — Daya Pernas: Core thesis for why AI changes cybersecurity from identity attacks toward vulnerability/exposure management

Implications: The episode suggests the next winners may be less obvious than the dominant AI platforms: power, networking, cyber exposure management, and select fintech/software names. For investors, valuation discipline and differentiated research matter more as AI reshapes who captures economics.