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, ALAB, AMD, AMPX, AMZN, APLD, APP, ASTS, AUR, BBAI, BFLY, BKSY, BRK.B, COHR, CRDO, ESTC, ETH, GOOG, GOOGL, HURA, HUT, IBIT, INOD, IONQ, JOBY, KTOS, LITE, MBLY, META, MSFT, MU, NBIS, NEE, NNOX, OKLO, OUST, PDYN, PL, PLTR, POET, QBTS, QUBT, RCAT, RGTI, RKLB, RLAY, RMBS, RR, SDGR, SGOV, TEM, USMV, VGT, VICR, VRT, VST.
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 September 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.
TeraWulf Inc. (WULF)
ai
cloud
energy
crypto
Synopsis
The upside case is a transition from miner optics to contracted AI campus owner. Strong returns are plausible if
Lake Mariner execution becomes repeatable and Kentucky scales with limited parent-level
dilution.
Thesis
TeraWulf can grow into a much larger
AI infrastructure owner if it keeps converting scarce power-backed campuses into long-duration lease revenue and shifts financing toward project-level structures, but the upside is governed more by capital formation and delivery speed than by customer demand.
Last Economy Alignment (0.7/1.0)
WULF owns a real AI bottleneck: powered,
interconnection-ready campuses. Cheaper cognition raises demand for its product rather than commoditizing it, though financing and permitting still cap value capture.
Critique
If scarce campuses become easier to finance or replicate than bulls expect, rent premiums can compress and WULF stays a capital-hungry builder despite low software disintermediation risk.
Cerebras Systems Inc. (CBRS)
semiconductors
ai
cloud
hardware
enterprise
Synopsis
A genuine architecture advantage and huge contracted demand create a path to a much larger recurring
AI infrastructure business. The key question is whether capacity delivery and customer diversification keep pace before larger clouds compress the economics.
Thesis
If it turns
wafer-scale speed into reserved-
throughput contracts, partner distribution, and higher-trust enterprise controls, Cerebras can become a recurring AI utility rather than a niche accelerator vendor; the equity compounds only if power-backed capacity and customer diversification arrive before fast
inference pricing compresses.
Last Economy Alignment (0.8/1.0)
Cerebras sells a scarce AI-era input: low-latency compute plus access to power-backed capacity. It is strongly helped by exploding
inference demand, but it does not fully control power, customers, or cloud distribution.
Critique
If fast
inference becomes an interchangeable backend inside larger clouds, OpenAI-compatible access and giant-customer bargaining power could compress pricing before Cerebras earns back its capacity buildout.
Space Exploration Technologies Corp. (SPCX)
space
communications
defense
ai
cloud
Synopsis
The equity case rests on turning launch cadence, satellite capacity, secure government access, and AI compute into more recurring, higher-trust contracts. Strong execution can still more than double value by 2031, but only if revenue quality rises faster than capital intensity.
Thesis
SpaceX should keep compounding because it owns AI-era bottlenecks that are hard to copy—launch cadence, orbital network capacity, spectrum, government trust, and growing compute—but from a near-$2T starting value the next leg depends on converting extreme
capex into denser recurring contracts rather than just adding more hardware.
Last Economy Alignment (0.8/1.0)
As AI scales, scarce physical bottlenecks become more valuable. SpaceX controls launch, spectrum, secure procurement channels, and capacity-backed compute, so it should gain from cheaper cognition; the main risk is not software commoditization but whether huge
capex earns durable returns before regulation or capacity cycles compress value.
Critique
If
Starlink stays mostly a bandwidth utility and AI becomes rented compute with little workflow control, agents and open models can capture the software layer while SpaceX absorbs the
capex, pushing pricing lower and leaving shareholders with an expensive telecom-and-launch asset.
NVIDIA Corporation (NVDA)
semiconductors
ai
hardware
networking
software
Synopsis
The core question is no longer whether AI demand exists, but how much of each AI deployment this company can own from silicon through site readiness. The business can still grow hard from here, yet its sheer size and premium expectations make stock upside more disciplined than explosive.
Thesis
NVIDIA can still roughly double by 2031 if it remains the default full-stack
AI factory standard and captures more of each deployment through systems, networking, software qualification, financing and power-site enablement, even as some accelerator share leaks to
custom silicon.
Last Economy Alignment (0.9/1.0)
It owns the core compute standard and is moving into deployment bottlenecks, so cheaper cognition mostly drives more volume through NVIDIA’s stack.
Critique
If
hyperscalers make workloads more portable and route more
inference to
custom chips while China stays closed, NVIDIA may stay essential but lose pricing power and see only modest equity upside from here.
SK hynix Inc. (SKHY)
semiconductors
hardware
ai
enterprise
Synopsis
The key question is whether scarce high-bandwidth memory stays scarce long enough for today’s profits to become a structurally higher earnings floor. If execution holds, the shares can compound well above a normal memory cycle, but the upside still depends on physical ramps rather than software-style economics.
Thesis
SK hynix can turn
HBM scarcity into a structurally better franchise if it keeps winning yield, packaging and customer qualification, using balance-sheet strength and customer lock-in to convert a cyclical memory business into a higher-quality
AI infrastructure supplier by 2031.
Last Economy Alignment (0.7/1.0)
Cheaper cognition increases AI memory demand, and SK hynix owns scarce
HBM, packaging and qualification loops; commodity memory exposure keeps it below the top band.
Critique
This may still be a scarcity-premium hardware story: if
HBM becomes broadly dual-sourced and buyers regain pricing power, process know-how and
LTAs may not stop margin and
multiple compression.