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

Dean and Deiya Pernas of Pernas Research, which claims an audited track record of over 1300 percent gross returns since 2017, roughly double the S&P 500, argue that the most important and misunderstood dynamics in markets right now center on data center power infrastructure, cybersecurity, and cross-border payments.

On AI broadly, Dean frames the ecosystem as four layers: chips, cloud infrastructure, frontier models, and applications. He argues open weight models are net bullish because commoditizing one layer increases demand across all others. Enterprises no longer paying roughly ten dollars per million input tokens to Anthropic or OpenAI will redirect that spending toward GPU compute and hyperscaler services. Deiya adds that hyperscalers now trade at roughly ten times EV to sales and are existentially compelled to make large capital outlays with uncertain returns, which creates attractive economics for companies receiving that spending without having to match it themselves.

The dominant theme Dean identifies in data center infrastructure is energy, specifically the bring your own power dynamic. U.S. electricity demand has been flat at roughly 4000 terawatt hours for 20 years, but AI is projected to consume 20 percent of the entire U.S. grid by 2030, approximately 100 gigawatts. Transmission queues are already running 18 to 24 months. Dean expects large monolithic one to three gigawatt data centers to hit a ceiling due to grid constraints and community opposition, with the industry shifting toward distributed one gigawatt sites that must communicate with each other, expanding bandwidth requirements by roughly 15 times.

Two specific power names Dean discusses are Capstone Energy and Stabilis Solutions. Capstone makes micro turbines from 50 kilowatts to a few megawatts with lead times of one to three months versus 18 to 24 months for larger turbine makers like GE Vernova, and was trading at roughly three times revenue compared to Bloom Energy at 15 to 20 times. Capstone has a fully diluted market cap of approximately 300 to 350 million dollars, hit its first year of profitability last year, and can produce roughly one gigawatt annually at full capacity. Stabilis liquefies natural gas, transports it via cryogenic trucks, and vaporizes it for data centers without pipeline access. It has a market cap of roughly 75 million dollars, recently signed a contract worth a couple hundred million dollars over two years, and counts SpaceX as a major customer. Pernas Research rates it speculative at one to three percent portfolio sizing.

On cybersecurity, Deiya argues that while roughly 80 percent of recent attacks have been identity-based, the threat is shifting toward AI-enabled software vulnerability exploitation, which lowers the cost for bad actors as model prices fall. She views the broad sell-off in cybersecurity stocks as a buying opportunity and identifies Tenable as the standout name. Tenable evolved from vulnerability scanning into exposure management in 2022, has a partnership with Anthropic, and trades at approximately four times EV to sales versus over 20 times for CrowdStrike and Palo Alto. It is just beginning to generate GAAP operating profit and has a visible upsell opportunity converting existing customers to its newer platform.

On software more broadly, Deiya notes that production-ready code costs have fallen 90 to 95 percent due to AI but no large enterprise software company has yet had its products supplanted. She argues defensible SaaS businesses have unique data, compliance requirements, deep integrations, or real-world components. She also flags that SBC normalization as a percentage of revenue could meaningfully improve GAAP earnings for many names. Dean separately argues Upwork is misunderstood relative to Fiverr, with weakness concentrated in sub-500 dollar jobs while higher-value engagements hold up, and the stock trades at just over one times EV to sales and roughly six times EV to free cash flow after accounting for SBC.

In payments, Deiya frames cross-border transfers as a durable multi-year opportunity. Traditional correspondent banking charges roughly five percent on average while fintechs like Wise and Remitly have brought costs closer to two percent by bypassing that system. Remitly focuses on migrant remittances with send volumes growing 30 to 40 percent annually and revenues growing roughly 25 percent, with brand trust helping defend its take rate. Wise operates at larger scale and is deliberately lowering its take rate to become a cost leader, building infrastructure that plugs directly into central banks and threatens correspondent banking and card networks, though over 90 percent of Remitly transactions still run on Visa and Mastercard rails. Paysafe is presented as exceptionally cheap with exposure to gaming and prediction markets that Stripe and Adyen avoid, carrying 2.5 billion dollars in debt but with a digital wallets subsidiary that alone could be valued at nine to ten times earnings and represents 60 to 70 percent of enterprise value. The speakers view stablecoins as still largely a crypto speculation vehicle with no meaningful adoption for everyday payments, making correspondent banking the most likely loser if stablecoins eventually gain traction in treasury functions.