Cloud & Capital · Market Tape
Market Tape is strategic decision support for technology-finance review. It connects public-market conditions, cloud benchmarks, and major AI infrastructure announcements without turning them into individualized investment advice or workload-level operating mandates.
The tracked universe contains 48 unique instruments displayed as 52 rows across the dashboard's market, macro, sector, and global views. A ticker shown in more than one useful category is counted once in status calculations and in the tracked-instrument total. The broader screened universe supplies the momentum leaderboard; only its five leaders and five laggards are displayed.
The upstream pipeline is scheduled to check for new data every 30 minutes during its weekday window, though GitHub Actions may delay or skip scheduled runs. On weekends and market holidays, the timestamp identifies the latest completed US trading session—not the page view, build, or deployment time.
The 0–100 exposure score is the share of unique tracked instruments with positive one-month relative strength versus SPY. Lower readings are Defensive, 40–60 is Hold/Neutral, and higher readings are Risk-On. These bands frame review priorities; they do not prescribe a portfolio allocation or cloud purchase. Organization-specific decisions still depend on workload demand, utilization, contract terms, budget, and business priorities.
Prices, momentum scores, sector rotation, and macro indicators are fetched from yFinance via a data pipeline scheduled to check every 30 minutes during weekday market hours and publishes to github.com/cloudandcapital/market-tape. Pipeline runs may publish a timestamp-only snapshot commit even when market values are unchanged. The frontend revalidates against the latest published snapshot every 30 minutes.
Macro & Market Internals
SPY, ^VIX, TLT, HYG, GLD, USO, IWM, QQQ, DXY — prices, 1M relative strength, trend grades, breadth metrics
US Sector ETFs
XLK, XLF, XLV, XLE, XLI, XLY, XLP, XLU, XLB, XLRE, XLC — ranked by 1-month relative strength vs SPY
Valuation Multiples (availability varies)
Approximate NTM P/S computed as: TTM P/S ÷ (1 + trailing revenue growth). Equal-weighted median across each basket. Labeled with ~ to communicate approximation. Yahoo Finance basket requests are cached for 30 minutes. If a basket cannot be fetched, its current multiple is shown as Unavailable; the archived April 24, 2026 quarterly fallback is not presented as current market data. The valuation API reports source and data date separately for each basket.
Public Cloud hyperscaler basket
AMZN · MSFT · GOOGL · ORCL
Blended corporate P/S — AWS, Azure, and GCP are segments within these companies, not isolated. Treat as a proxy.
SaaS basket (representative BVP Cloud Index names)
CRM · NOW · SNOW · DDOG · ZS · HUBS · WDAY · VEEV
Eight liquid names from the BVP Nasdaq Emerging Cloud Index. Not the full index; WCLD ETF is the closest tradeable equivalent.
AI infrastructure semiconductor basket
NVDA · AVGO · AMD · MU · MRVL · CBRS
Semiconductor and AI hardware names. Dominated by NVDA; results are sensitive to NVDA revenue forecasts. CBRS (Cerebras, IPO May 2026) added Jun 2026 — wafer-scale inference chips, extreme P/S outlier; median calculation isolates it from distorting the basket.
This major-deal tracker is compiled manually from public announcements and press releases. Each row is sourced and classified individually: Signed means an executed agreement is confirmed; Announced means the parties publicly described a deal or partnership without equivalent confirmation of execution; Target means a planned capacity or investment goal; and Reported / in talks means a negotiation or third-party report that is not a completed commitment. Dollar values and capacity are not aggregated across unlike statuses or agreement types. The signed-dollar headline includes only structured, company-disclosed, comparable signed compute/cloud-service values; equity and reported or estimated figures are excluded. The data is static until a material update — source code at lib/aiCompute.ts.
OpenAI — AWS (Trainium)
Signed · compute/cloud service · value: $138B (company-disclosed) · capacity: Approximately 2 GW Trainium in expanded agreement (company-disclosed) · Expansion over 8 years · announced Feb 27, 2026
OpenAI says the existing $38B AWS compute agreement was expanded by $100B, not replaced by it. $138B combines the original agreement and incremental expansion once; Amazon’s separate $50B equity investment in OpenAI is excluded. Do not add the 2 GW expansion capacity to unrelated infrastructure targets.
OpenAI — Firmus (Malaysia AI factories)
Announced · compute/cloud service · value: Undisclosed (undisclosed) · capacity: OpenAI allocation undisclosed (undisclosed) · Multiyear · announced Sep 8, 2026
Firmus and OpenAI announced a partnership under which OpenAI will contract capacity at two Malaysian sites. Firmus’s more than 900 MW is its entire contracted customer portfolio, not OpenAI-specific capacity; the release does not state that this particular contract has been executed.
Amazon — Qualcomm (custom silicon / optical)
Signed · hardware/chip partnership · value: Undisclosed (undisclosed) · capacity: Undisclosed (undisclosed) · Multigenerational · announced Sep 8, 2026
The SEC filing confirms an executed warrant and initial purchase commitments, but does not disclose their firm dollar value. Up to $60B is the warrant-linked maximum payment threshold, not a firm order or compute contract value; it is excluded.
Australian AI ecosystem — NVIDIA Cloud Partners / DSX
Target · infrastructure target · value: Undisclosed (undisclosed) · capacity: Up to 2 GW by 2027 (ecosystem target) (company-disclosed) · Target by 2027 · announced Sep 9, 2026
Capacity is a multi-provider Australian infrastructure buildout target, not a signed customer allocation or additive contract value.
Unidentified customer — SpaceX / xAI
Reported / in talks · compute/cloud service · value: $1.11B/month reported (run rate; total undisclosed) (reported) · capacity: Undisclosed (undisclosed) · Undisclosed · announced Sep 2026
Counterparty and total contract value are unidentified. The monthly reported run rate is not annualized or included in the signed-dollar headline.
Anthropic — AWS (Trainium / Inferentia)
Signed · compute/cloud service · value: $100B+ (company-disclosed) · capacity: Up to 5 GW (company-disclosed) · 10 years · announced Apr 2026
Anthropic — Google Cloud / TPU (Broadcom)
Signed · compute/cloud service · value: Undisclosed (undisclosed) · capacity: Multiple gigawatts (company-disclosed) · Deployment starts 2027 · announced Apr 2026
The primary does not disclose the previously shown $200B, 3.5 GW, or five-year term, so those figures have been removed.
Anthropic — xAI / SpaceX (Colossus 1)
Signed · compute/cloud service · value: ~$40B reported (reported) · capacity: More than 300 MW / 220,000+ NVIDIA GPUs (company-disclosed) · 3 years reported · announced May 6, 2026
Anthropic confirms the signed agreement and capacity. The approximately $40B value and three-year term remain third-party reported and are excluded from the company-disclosed signed-dollar headline.
Anthropic — Nscale (West Virginia / Vera Rubin)
Reported / in talks · compute/cloud service · value: ~$45B reported (reported) · capacity: ~460 MW reported (reported) · 6 years reported · announced Aug 26, 2026
Reuters reported the West Virginia agreement and Vera Rubin deployment from a person familiar with the matter; Anthropic declined to comment. Value, term, and capacity are excluded from the signed-dollar headline.
Anthropic — Lambda (Texas / Nueces County)
Reported / in talks · compute/cloud service · value: ~$35B reported (reported) · capacity: ~350 MW reported (reported) · Undisclosed · announced Aug 31, 2026
Reuters reported that Anthropic signed for Lambda cloud capacity associated with a Nueces County, Texas data center and described Lambda as NVIDIA-backed. No first-party confirmation naming the parties was found; value, capacity, location, provider structure, and NVIDIA involvement remain reported and are excluded from the company-disclosed signed-dollar headline.
Anthropic — AMD (Instinct MI450 / Helios)
Announced · hardware/chip partnership · value: Undisclosed (undisclosed) · capacity: Up to 2 GW (company-disclosed) · Starts H1 2027 · announced Jul 22, 2026
The potential $5B equity investment is displayed separately and is not compute procurement.
Anthropic — Volta (Norway project)
Reported / in talks · compute/cloud service · value: ~$10B reported (reported) · capacity: ~133 MW reported (reported) · 6 years reported · announced Aug 4, 2026
No first-party source publicly naming Anthropic was found; every figure is reported.
OpenAI — Stargate (Oracle, SoftBank, partners)
Target · infrastructure target · value: Undisclosed current value (undisclosed) · capacity: Original 10 GW goal surpassed (company-disclosed) · Ongoing · announced Jan 2025 (current update 2026)
The 10 GW figure is retained only as the surpassed historical goal, not current capacity.
OpenAI — NVIDIA (Vera Rubin)
Signed · hardware/chip partnership · value: Undisclosed (undisclosed) · capacity: 3 GW dedicated inference + 2 GW training (company-disclosed) · — · announced Feb 27, 2026
OpenAI directly reports securing 3 GW of dedicated inference capacity and 2 GW of training capacity on NVIDIA Vera Rubin systems; compute dollar value is undisclosed. NVIDIA's $30B equity stake in OpenAI (part of the $110B Feb 2026 funding round) is classified separately and does not enter the compute total.
Meta — NVIDIA (Blackwell + Vera Rubin + Grace)
Announced · hardware/chip partnership · value: Undisclosed (undisclosed) · capacity: Millions of Blackwell and Rubin GPUs (company-disclosed) · Multiyear · announced Feb 2026
This hardware supply layer is distinct from Meta cloud-service contracts; the infrastructure descriptions are not additive.
Meta — AMD (Instinct MI450+)
Signed · hardware/chip partnership · value: Undisclosed (undisclosed) · capacity: Up to 6 GW (company-disclosed) · Multiyear · announced Feb 24, 2026
Meta — Amazon (Graviton)
Signed · compute/cloud service · value: Undisclosed (undisclosed) · capacity: Tens of millions of Graviton cores (company-disclosed) · Multiyear · announced 2026
Meta — CoreWeave (Vera Rubin)
Signed · compute/cloud service · value: ~$21B (company-disclosed) · capacity: AI cloud capacity; no comparable GW disclosed (undisclosed) · Through Dec 2032 · announced Apr 9, 2026
Cloud-service contract using NVIDIA systems; distinct from Meta–NVIDIA’s hardware partnership and not additive at the infrastructure layer.
Jane Street — CoreWeave
Signed · compute/cloud service · value: ~$6B (company-disclosed) · capacity: Undisclosed (undisclosed) · Undisclosed · announced Apr 15, 2026
OpenAI — Cerebras Systems
Announced · compute/cloud service · value: Undisclosed (undisclosed) · capacity: 750 MW (company-disclosed) · Through 2028 · announced Jan 14, 2026
OpenAI — SB Energy / NVIDIA (PORTS-Pike, Ohio)
Signed · compute/cloud service · value: Undisclosed (undisclosed) · capacity: 4.25 IT-GW initial; approximately 3.75–3.8 IT-GW optional; OpenAI describes the overall agreement as approximately 8 IT-GW (company-disclosed) · 20-year lease, phased availability beginning 2028 · announced Aug 17, 2026
NVIDIA's $105B figure is a conditional residual-value guarantee, not compute value or committed spend; it is not reflected in the compute value field. NVIDIA's $1.5B investment in SB Energy is a separate equity transaction. This row is not merged into the existing OpenAI–NVIDIA Vera Rubin row because overlap between the two agreements has not been ruled out.
Google — SpaceX / xAI
Reported / in talks · compute/cloud service · value: ~$30B reported (reported) · capacity: Undisclosed (undisclosed) · 32 months reported · announced Jun 5, 2026
Slow-moving external statistics maintained manually in lib/industryBenchmarks.ts. Each entry records its last update, next review date, and review cadence. The automated freshness check warns within 14 days of a due date and fails once a benchmark is overdue. Run npm run check-benchmarks to see current freshness status.
Updated Oct 1, 2026 · Next review Nov 1, 2026 · monthly
October 1, 2026 dedicated comparisons: H100 $2.99–$10.98/GPU-hour; H200 $3.99–$10.85; B200 $5.99–$16.11. Availability language differs by provider and purchase model, so the benchmark does not characterize any generation as uniformly available or supply-constrained. Provider, commitment, marketplace, and configuration differences make rental ranges non-comparable to hardware purchase prices. NVIDIA Q2 FY2027 is retained separately only for NVIDIA-specific demand and product-ramp commentary.
Updated Aug 31, 2026 · Next review Feb 28, 2027 · semi-annual
CBRE H1 2026 primary-market figures: net absorption 1,456.2 MW (+11.7% YoY), under construction 7,481.1 MW (+24.8%), and total supply 10,903 MW (+33.7%). Capacity remains tight despite construction growth: vacancy reached a record-low 1.4%, 80.4% of construction was preleased, and less than 1,500 MW remained available. The prior construction-decline explanation is no longer current.
Updated Apr 24, 2026 · Next review Nov 1, 2026 · semi-annual
Peak NTM revenue multiple reached by the SaaS cohort during the 2021 bull market. Used as a compressed-from reference point. This is a historical data point and will not change. Review cadence is semi-annual only to confirm the narrative framing is still accurate.
Updated Sep 13, 2026 · Next review Nov 1, 2026 · quarterly
Full-year 2026 guidance after Q2 earnings: Amazon approximately $220B; Microsoft approximately $175B (lease classification shifted); Alphabet $195–205B; Meta $130–145B. Oracle Q1 FY2027 cloud infrastructure revenue was $7.4B (+121% YoY); RPO $664B, with more than $30B of new AI cloud bookings, 850 MW of added data-center capacity and over 300,000 GPUs delivered since Q4. Those Oracle aggregates are not individual customer compute contract values. Google plans at least €13B of Finland digital-infrastructure investment during 2027–28, with a 22-year nuclear PPA and other energy support. AWS and NVIDIA plan 2 million additional NVIDIA GPUs in 2027–28. All are capex/capacity or earnings context, not signed customer-deal totals. Values: Expanding / Stable / Contracting. Recheck after the next earnings cycle.
Updated Aug 31, 2026 · Next review Nov 25, 2026 · quarterly
NVIDIA Q2 FY2027 (quarter ended July 26, 2026). Total revenue was $96.2B (+106% YoY, +18% QoQ); Data Center revenue was $89.0B (+117% YoY, +18% QoQ). Q3 FY2027 total-revenue guidance is $108.0B ±2%. Vera Rubin is in full production. NVIDIA did not disclose a directly comparable standalone networking figure under the new presentation, so none is shown.
Lumen is an AI-generated synthesis of the market data and sourced benchmarks displayed on Market Tape. Its role is to surface relationships and review priorities, not to create new facts. Completed analysis is shared within the market-data refresh window so visitors see a consistent reading of the same snapshot.
Market Tape provides strategic context, not individualized investment, financial, tax, procurement, or workload advice. Organization-specific decisions still require workload demand, utilization, contract terms, budget, risk tolerance, and business priorities. Data may be delayed, revised, approximate, or temporarily unavailable; follow the linked primary sources before acting on a benchmark or announcement.