Week of June 01, 2026 — a synthesis of 7 daily investment briefs and the underlying blended pricing data.
The week's most kinetic move was MI25, which closed 41.4% lower at $0.25/hr blended. Across all instruments, the median GPU rental rate declined 2.2% week-on-week.
Week in Briefs
Across 7 daily intelligence briefs this week, the analysis surfaced 39 investable hypotheses. Here's what built conviction.
Week at a Glance
| Date | Headline Hypothesis |
|---|---|
| Mon Jun 1 | The H200 Cliff Is 30–60 Days From Hitting Core Markets — Front-Run the Margin Compression in Neocloud Equities |
| Tue Jun 2 | H100 Spot Is Being Displaced by H200, But Divergence Is Geographically Bifurcated — Creating a Durable Regi… |
| Wed Jun 3 | The H100 SPOT Floor Has Passed — A Stealth Reflation is Underway |
| Thu Jun 4 | The Montreal H100 SPOT Breakout Is a Leading Indicator of Sustained North American Capacity Tightening |
| Fri Jun 5 | The Montreal–Stockholm SPOT Squeeze Is a Leading Indicator of a Wider H100 Capacity Tightening Cycle |
| Sat Jun 6 | AWS ASIC Spot Repricing Signals Inference-at-Scale Inflection — Ahead of AVGO/Custom Silicon Re-Rating |
| Sun Jun 7 | The H100 Spot Discount is Structurally Compressing — And the Vera Rubin Delivery Is the Only Relief Valve |
Convictions That Built This Week
H100 — appeared in 6 briefs (Mon Jun 1, Tue Jun 2, Wed Jun 3, Thu Jun 4, Fri Jun 5, Sat Jun 6)
| Rank | Hypothesis | Conviction | Horizon | Key Catalyst | |---|---|---|---|---| | #1 | Montreal/Stockholm H100 recovery propagates globally | High | 4–6 weeks | London/Tokyo SPOT delta | | #2 | Power-secured markets earn permanent greenfield premium | High | 6–18 months | Quebec/Nordic capacity filings | | #3 | Inferentia scarcity = ASIC inference structural tightening | Medium | 2–4 weeks | APAC Inf2 discount compression | | #4 | CRWV valuation is fragile to H100 pricing normalization | Medium | 2–3 months | Aug 12 earnings + H200 SPOT emergence | | #5 | Cooling IP becomes Blackwell
Framing across the week: CoreWeave's H100 1-Year Reserved Contracts Are Now Underwater — A Hidden Credit Event Is B; The Budget Inference Tier Is Real — Legacy GPU Catalog Activation Signals a New $0.07–1.23; H100 Spot Is Being Displaced by H200, But Divergence Is Geographically Bifurcated — Creati
Inferentia — appeared in 4 briefs (Wed Jun 3, Thu Jun 4, Fri Jun 5, Sat Jun 6)
| # | Thesis | Key Ticker Evidence | Equity Expression | Timeframe | Conviction | |---|---|---|---|---|---| | 1 | ASIC Inference Demand Inflection → AVGO re-rate | INFERENTIA Mumbai +897%, TRAINIUM Ohio +284% | Long AVGO on guidance-miss dip | 6-12 weeks | High | | 2 | H100 Montreal Spot Floor Reset → CRWV recovery | H100 Montreal +122% 30d, SPOT/Reserved gap closing | Long CRWV before Aug 12 earnings | 4-10 weeks | High | | 3 | SPOT/On-Demand Spread Compression → NVDA re-rate | H100 SPOT rising flat against on-demand ceiling | Long NVDA on macro-driven dip | 8-12 weeks | Medium-High | |
Framing across the week: Broadcom's ASIC Earnings (Today) Will Accelerate the GPU-to-ASIC Rotation Thesis — and Val; AWS Is Deliberately Restricting Inferentia/Trainium SPOT Supply — This Is a Monetization S; The SPOT/ON_DEMAND Inversion in Inferentia Is the First Sign of a Broader AWS Custom Silic
Google — appeared in 3 briefs (Wed Jun 3, Thu Jun 4, Sun Jun 7)
The $40B Google Texas announcements and OpenAI Stargate are real capital but face a gauntlet of permitting, grid connection, and political opposition that will stretch timelines by 3–7 years. JLL projects 100GW of new datacenter development 2026–2030 — but the IEA says consumption doubles to 945 TWh, and the University of Michigan's finding that renewables can't meet uptime requirements means gas generation (which is politically and regulatorily constrained) is the only baseload option.
Framing across the week: The Neocloud-to-Hyperscaler Pipeline Trade — Converted Miners Are the Short-Cycle Supply S; _Power-Secured Existing Operators Are the Only Safe Long in AI Infrastructure — Greenfield _; The SpaceX-Google GPU Deal Is a Pricing Shock for AWS/Azure — And Neocloud Premium Compres
Open Questions Into Next Week
- AWS ASIC Spot Repricing Signals Inference-at-Scale Inflection — Ahead of AVGO/Custom Silicon Re-Rating (raised Jun 6)
- H100 Montreal Is the Canary for a North American Spot Floor Reset (raised Jun 6)
- The H100 Reserved/Spot Spread is Encoding a Hidden Supply Ceiling Signal (raised Jun 6)
- The H100 Spot Discount is Structurally Compressing — And the Vera Rubin Delivery Is the Only Relief Valve (raised Jun 7)
Get the Weekly Pulse
This analysis is part of Signwl's weekly research. Subscribe free — Tuesday delivery, no spam.
Market Snapshot
Blended GPU compute costs fell 2.2% this week across major cloud providers, led by MI25's 41.4% decline.
Top Movers
| GPU | Blended Price | WoW Change | Class |
|---|---|---|---|
| MI25 | $0.25/hr | ▼ 41.4% | General |
| A10 | $0.92/hr | ▼ 9.8% | Inference |
| ALVEO_U30 | $0.21/hr | ▲ 9.4% | General |
| TRAINIUM | $0.67/hr | ▼ 8.8% | Training |
| V710 | $0.89/hr | ▼ 8.5% | General |
| RTX PRO 6000 | $0.52/hr | ▼ 8.3% | General |
| M60 | $0.28/hr | ▼ 5.1% | Inference |
Blended pricing = average of spot, on-demand, and 1-year reserved rates across major cloud providers.
Training vs Inference
Training-class GPU pricing fell this week (avg $4.25/hr, -1.0% WoW), while inference-class pricing fell (avg $0.52/hr, -1.8% WoW).
The training-to-inference price ratio stands at 8.2x — widening compared to last week. The elevated spread suggests strong demand for training compute relative to inference, consistent with ongoing large model training activity.
Regional Spotlight: Europe
Europe trades at a 5% discount to global averages this week, with 34 GPU types available across the region. The most expensive GPUs in the region are GB200 ($14.79/hr), H200 ($10.76/hr), TPU_V7 ($10.43/hr).
For detailed pricing data across all Europe sub-regions, see the full regional profile.
Implications
For cloud buyers: Asia Pacific continues to offer the lowest average GPU pricing ($2.02/hr blended average). For workloads with regional flexibility, the gap between Asia Pacific and Middle East is $2.07/hr — a 103% premium. Compare regional pricing →
For semiconductor analysts: H100 pricing softness alongside MI300X strength suggests potential AMD share gains — worth monitoring ahead of next earnings. Blackwell (B200) blended pricing at $7.65/hr (-3.1% WoW) provides an early read on next-generation adoption curves. View all GPU profiles →
For GPU investors: Declining rental rates are compressing returns — monitor payback period shifts closely. Model scenarios with the GPU ROI Calculator →