Disclaimer: This analysis is for informational and educational purposes only and does not constitute investment advice. All investments carry risk, including the risk of loss. Past performance does not guarantee future results. Please consult with a qualified financial advisor before making investment decisions.
Position Disclosure: The author holds positions in the following symbols: AAOI, AAPL, ACHR, ALAB, AMD, AMPX, AMZN, APLD, APP, ASTS, AUR, AVAV, AVGO, BBAI, BFLY, BKSY, BRK.B, CEG, COHR, CRDO, CRWV, ESTC, ETH, GOOG, GOOGL, HURA, HUT, IBIT, INOD, IONQ, IREN, JOBY, KTOS, LITE, LMND, MBLY, META, MSFT, NBIS, OKLO, OUST, PDYN, PL, PLTR, POET, QBTS, QUBT, RCAT, RGTI, RKLB, RLAY, RMBS, RR, SDGR, SERV, SGOV, SITM, SMR, SOUN, SYM, TEM, USMV, VGT, VICR, VRT.
Every week, we score ~120 public companies on one question: as AI changes how businesses think, build, and compete, who benefits most? These five ranked highest this week — a sample of the full analysis available to members.
How We Picked Them: Selected from our full coverage based on growth potential, AI advantage, and how well-positioned each company is as of May 15, 2026. We diversified across sectors so you see a range of opportunities, not just one hot corner of the market.
Also available as a PDF download.
Nebius Group N.V. (NBIS)
ai
cloud
enterprise
software
hardware
Synopsis
The upside is real if scarce power converts into utilized clusters and the
inference layer becomes economically sticky. The stock already discounts much of the near-term ramp, so execution matters more than narrative from here.
Thesis
Nebius has real non-linear upside because it controls scarce AI power, GPU capacity and a growing
inference stack, but the stock will only compound if management turns contracted capacity into highly utilized revenue fast enough and proves
Token Factory can lift it above commodity compute pricing.
Last Economy Alignment (0.7/1.0)
Nebius benefits directly from AI demand because it owns scarce power-and-compute control points, but its usage-priced cloud economics can still compress if
hyperscalers outbuild it or software differentiation stays thin.
Critique
This may still be a peak-scarcity compute rental story: if
hyperscalers fund faster, customers buy capacity directly, and Nebius’s
inference and trust layers remain features rather than control points, revenue can rise while pricing, returns and valuation compress.
Applied Digital Corporation (APLD)
ai
cloud
energy
crypto
Synopsis
This is a leveraged buildout of scarce AI data-center capacity where demand looks real and contracted revenue is increasingly visible. The upside now depends more on financing discipline and
energization credibility than on another demand narrative.
Thesis
Applied Digital can grow into a materially larger AI infrastructure owner if it converts scarce powered sites and 900
MW of contracted campuses into recurring lease revenue faster than financing costs, utility gates, and
dilution erode common-equity value; the upside is real because power is scarce, but the stock now needs execution more than another narrative
rerating.
Last Economy Alignment (0.7/1.0)
APLD controls scarce powered campuses and long-duration capacity contracts that become more valuable as AI demand rises. It benefits directly from AI scaling, but it still depends on utilities and outside capital to turn
megawatts into shareholder value.
Critique
If outsourced AI campuses are mostly finance-heavy shells, value shifts to
hyperscalers and lower-cost-capital landlords while APLD keeps construction, power-timing, and
dilution risk; then revenue can rise without strong per-share value creation.
Cerebras Systems Inc. (CRBS)
semiconductors
ai
cloud
hardware
enterprise
Synopsis
A differentiated
inference stack can compound into a large AI infrastructure business if promised capacity becomes real, utilized capacity. The stock can still work, but much of the easy upside was pulled forward on day one.
Thesis
Cerebras has a credible path from bespoke AI hardware vendor to multibillion-dollar
inference utility because it controls differentiated
wafer-scale compute and has real OpenAI and AWS distribution vectors; the business could scale non-linearly, but the stock starts from a valuation that already prices in years of success.
Last Economy Alignment (0.8/1.0)
Cerebras owns a real AI bottleneck: fast
inference hardware plus deployed capacity. That is highly aligned with rising AI usage, though
hyperscaler bargaining power and capital intensity cap the score.
Critique
If fast
inference is a feature rather than a durable platform,
hyperscalers and GPU vendors can commoditize the API layer, squeeze pricing, and leave Cerebras with concentrated demand and an expensive build-out.
CoreWeave, Inc. (CRWV)
cloud
ai
software
enterprise
Synopsis
The upside case rests on converting a massive
backlog and power pipeline into live clusters faster than bigger rivals. The swing factor is whether financing innovation and
inference attach outrun
dilution and multiple compression.
Thesis
CoreWeave can still be a strong 5-year winner if it keeps turning signed AI demand,
contracted power, and structured finance into live high-
utilization clusters, then adds more
inference and workflow revenue before raw GPU scarcity normalizes.
Last Economy Alignment (0.8/1.0)
CoreWeave sells a primary AI bottleneck: delivered compute and power capacity, then wraps it in software that makes those resources usable at scale. That is strongly aligned with the
Last Economy, but cheaper capital at
hyperscalers and a future supply catch-up can cap value capture.
Critique
If AI capacity stops being scarce and larger clouds keep the cheapest capital, CoreWeave's software layer may not be thick enough to prevent utility-like pricing,
dilution, and lower equity returns.
Palo Alto Networks, Inc. (PANW)
cybersecurity
software
enterprise
ai
cloud
Synopsis
The core case is steady compounding, not hypergrowth. AI, identity, and security-operations expansion can lift
wallet share enough to roughly double value by 2031, but only if recent acquisitions become one coherent trust layer.
Thesis
PANW should remain an AI-era security compounder: as
machine identities, agent workflows, and attack volume grow, it can capture more
wallet share by owning policy, logging, identity, and response control points, pushing revenue into the low-$20B range by 2031 without requiring heroic
TAM assumptions.
Last Economy Alignment (0.7/1.0)
AI makes security more necessary, and PANW controls policy, logging, and identity layers that get stickier as machine actors multiply.
Critique
If
platformization is mostly
discount-led bundling and agents execute inside rival cloud or identity stacks, PANW may keep customers but lose pricing power and see its
premium multiple compress.