Companies

Amazon / AWS (NASDAQ: AMZN) sits at the centre of multiple Signwl theses: AWS is the principal B200 supplier whose regional spot dispersion drives the H100 depreciation arc, and AWS's own Neuron silicon (Trainium, Inferentia) is the cleanest spot-side evidence of custom-silicon substitution working in real time.

Anthropic appears in Signwl's briefs as the principal compute-buyer signal in the inference-side of the custom-silicon substitution arc, with the $21bn TPU order (Broadcom / Google, 2026–2027 delivery) as the single largest non-GPU inference infrastructure commitment yet disclosed.

Google (NASDAQ: GOOGL) is the cleanest publicly-positioned beneficiary of the custom-silicon substitution arc: TPU is now a multi-decade investment with named third-party customers (Anthropic $21bn order), and GCP's recent EU TPU v5 rollout shows up directly in Signwl's spot tape.

Meta (NASDAQ: META) sits in two simultaneous positions in Signwl's analysis: a Broadcom MTIA custom-silicon partner (positively framed in the substitution arc) and one of CoreWeave's two material customers (a transmission channel for the H100 depreciation reset).

Microsoft (NASDAQ: MSFT) appears in Signwl's briefs primarily through two angles: the Azure Maia / GB200 custom-silicon ramp (a Broadcom partner) and the role Microsoft plays as one of CoreWeave's two material customers — the latter being a transmission channel for the H100 depreciation reset.

OpenAI appears in Signwl's briefs as a multi-vendor compute customer — partnered with Broadcom on custom silicon, AMD on training infrastructure (MI450 / Helios), and Oracle / Crusoe on confirmed Abilene capacity. The breadth of partnerships makes OpenAI one of the cleanest indicators of compute supply chain diversification.

AMD (NASDAQ: AMD) appears across Signwl's briefs through two distinct lenses: the MI450 / Helios training-rack roadmap as a credible Blackwell challenger, and the CPU-side of the GPU-to-CPU ratio shift that AMD itself has guided toward (8:1 → 1:1) — a structural change with downstream implications for inference GPU demand.

Broadcom (NASDAQ: AVGO) is the consistent picks-and-shovels reference across Signwl's custom-silicon substitution analysis. The company supplies the chip-design, packaging, and networking IP behind multiple hyperscaler ASIC programmes (Google TPU, Meta MTIA, Anthropic, OpenAI), which positions it to capture revenue regardless of which specific ASIC wins share from Nvidia.

Intel (NASDAQ: INTC) appears in Signwl's briefs primarily as a counterpoint: the Gaudi accelerator line is being deprecated even faster than AWS Trainium (which is itself underperforming), strengthening the case that AI ASIC competition is harder than headlines suggest. CPU-side, Intel is the share-loser counterpart to AMD's EPYC gains.

NVIDIA (NASDAQ: NVDA) is the central name in Signwl's AI infrastructure analysis — H100 fragmentation, B200 demand-pull squeeze, custom-silicon substitution risk, and the H100 depreciation reset all run through Nvidia's spot tape and forward backlog.

TSMC (NYSE: TSM) is the chokepoint behind every name in the AI infrastructure complex: Nvidia's Blackwell, Broadcom's custom-silicon programs, AMD's MI series, Apple, Google TPU — all flow through TSMC's leading-edge nodes. Coverage in Signwl's briefs is light but consistently positive.

CoreWeave (NASDAQ: CRWV) sits at the intersection of three structural dynamics tracked by Signwl: H100 spot-market fragmentation, AWS custom-silicon substitution, and the power-as-binding-constraint thesis. The company appears in our analysis weekly because each of these themes interacts directly with its economics.

Crusoe Energy is a private AI infrastructure operator with a distinctive model: methane-to-power generation feeding owned-and-operated data centre capacity. It appears in Signwl's analysis as a core data point in the crypto-to-AI conversion arc and as a candidate discount-pricing actor in regional H100 markets.

Lambda Labs is a private GPU cloud provider with structural exposure to the same dynamics affecting CoreWeave: H100 spot fragmentation, power-cost pass-through risk, and the broader competition between rented and owned compute. Our coverage is currently light — Lambda has appeared as a comparison / candidate name in two recent briefs.

Core Scientific (NASDAQ: CORZ) sits at an unusual intersection in Signwl's analysis: it is positively framed as a power-rich asset within the crypto-to-AI conversion thesis, but also identified as a pair-trade short against CoreWeave because of the customer-concentration relationship.

HIVE Digital (NASDAQ: HIVE) appears in Signwl's briefs as a concrete example of how power-secured announcements are repricing the crypto-to-AI conversion cohort — the +25% move on the 320 MW Ontario plan is the cleanest market validation we have seen of the power-as-binding-constraint thesis.

Hut 8 (NASDAQ: HUT) is the second-most-anchored name in the crypto-to-AI conversion arc, distinguished by the $9.8bn 15-year lease at Beacon Point Texas with an investment-grade tenant — structurally a corporate-bond-equivalent revenue stream.

IREN (NASDAQ: IREN) is the most consistently positively-framed name in Signwl's analysis. It sits at the centre of the crypto-to-AI conversion arc and the power-as-binding-constraint thesis: a former Bitcoin miner with pre-permitted, grid-connected power at scale, now contractually committed to AI compute via the $3.4bn Nvidia deal.

TeraWulf (NASDAQ: WULF) appears in Signwl's briefs as part of the crypto-to-AI conversion cohort, sharing the structural attribute that defines the group: pre-permitted, grid-connected power capacity in markets where new interconnection takes 4+ years.

Constellation Energy (NASDAQ: CEG) is the largest US nuclear operator and appears in Signwl's briefs alongside Vistra and NRG in the pure-play power infrastructure long thesis. The June 1 earnings call is identified as a material catalyst for the broader power-constraint read.

Eaton (NYSE: ETN) appears in Signwl's briefs as a power-equipment OEM with the structural advantage of multi-year lead times (4+ years for transformers) translating to long revenue visibility on booked orders.

GE Vernova (NYSE: GEV) appears in Signwl's briefs as the representative power-equipment OEM benefiting from the multi-year capex cycle in transformers, gas turbines, and grid infrastructure driven by AI data centre demand.

Kinder Morgan (NYSE: KMI) is the second named natural gas pipeline operator in Signwl's gas-to-power thesis, with confirmed exposure via Liberty County, Texas data centre infrastructure.

Vistra (NYSE: VST) appears in Signwl's briefs as a pure-play power infrastructure operator with direct data centre exposure. It is consistently grouped with Constellation Energy and NRG in our "long bias" power-constraint framing.

Williams Companies (NYSE: WMB) appears in Signwl's briefs as a named pipeline operator with explicit data centre exposure, representing the gas-to-power capacity that increasingly bypasses constrained electrical grid interconnection.

Digital Realty (NYSE: DLR) appears in Signwl's briefs primarily through exposure to the Virginia data-centre cluster and the broader power-rich REIT framing. The FERC transmission-cost reallocation question is the principal near-term overhang.

Equinix (NASDAQ: EQIX) appears in Signwl's briefs as the canonical example of a data centre REIT with existing grid interconnection — an asset class re-rated upward by the broader power-as-binding-constraint thesis.

Nscale is a private Nordic AI infrastructure operator that has emerged as the cleanest leading indicator for European AI capacity buildout. The $790m financing structured as utility-grade accordion debt suggests pre-IPO positioning analogous to the early-stage IREN trajectory.