# Carmen Li argues in Silicon Data for separate H100 rental-rate indices [Canonical reading](https://escapemeridian.com/readings/carmen-li-argues-in-silicon-data-for-separate-h100-rental-rate-c317825d) Carmen Li’s article in Silicon Data was published June 16, 2026. She argues that H100 rental rates need separate Hyperscaler and Neocloud indices because the same chip’s two market tiers behaved differently over the April–May window. The Hyperscaler index moved +0.27% end to end, within a roughly 2% range; the Neocloud index moved +3.8%, with an 11.7% trough-to-peak swing. Li attributes the contrast to market structure: longer reservations and an administered price ceiling make the Hyperscaler rate slow to reprice, while shorter commitments and more price-sensitive buyers let Neocloud prices respond more quickly to supply and demand. She says the indices serve different purposes: the Hyperscaler series can anchor reservation pricing and budgeting, while the Neocloud series can inform month-to-month rental decisions and offer an earlier read on demand. Li’s conclusion is that blending the indices would average away the stability of one and the responsiveness of the other. She cautions that the series show price movements, not their cause. Whether Neocloud strength continues remains unknown; Li says June data may help answer that question. Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims. ## Original source Silicon Data · Article Original source date: Jun 16, 2026 (2026-06-16). Date basis: Date stated on the publisher’s page. Added to Escape Meridian: 2026-11-06T13:13:24.206Z. First observed: 2026-10-25T01:53:38.909Z. Retrieved: 2026-10-25T01:55:38.181Z. - [H100 Index, May 2026: Same Chip, Two Tiers, Two Regimes - Silicon Data — GPU Performance Data for Companies](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Limits Text as retrieved on October 25, 2026; originally published June 16, 2026. Li presents the structural explanations for the different price behavior as likely or plausible; the article says the index movement itself does not establish its cause. The article says June data may help answer whether Neocloud strength continues; that outcome is not established in this reading. ## Open question What will June data show about whether Neocloud strength continues? ## Source-stated Carmen Li argues that the two H100 rental-rate tiers behaved differently over the same April–May window. > Over those two months, the Hyperscaler index moved +0.27% end to end, inside a peak-to-trough range of about 2%. The Neocloud index moved +3.8% end to end, but over a far wider path: - [Supporting source](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Source-stated Li attributes the difference in price movement to distinct market structures, including longer reservations and a price ceiling for Hyperscalers versus shorter commitments in the Neocloud tier. > The Hyperscaler tier functions as a price ceiling that AWS, Azure, and Google Cloud post above a large book of already-priced reserved capacity. - [Supporting source](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Source-stated Li says the indices serve different uses: the Hyperscaler series as a budgeting reference and the Neocloud series for spot-exposed decisions and demand monitoring. > The Hyperscaler index suits structured reservation pricing and enterprise budgeting. Its stability is the point: - [Supporting source](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Source-stated Li argues that combining the two series would average away the distinct properties that make each useful. > Blending them into one number would average away the stability that makes the Hyperscaler series a budgeting anchor and the responsiveness that makes the Neocloud series an earlier read on demand. - [Supporting source](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Source-stated Li says the price series show the Neocloud move but do not establish its cause. > We note this as an interpretation; the series itself shows the move, not its cause. - [Supporting source](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Source-stated The article was written by Carmen Li and published June 16, 2026. > Written by Carmen Li > Founder at Silicon Data > \# Industry Jun 16, 2026 8 Mins Read - [Supporting source](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Source and contributor context Silicon Data · Publisher The publisher is identified here. No individual author has been inferred. ## Source media Media retrieval status: unchecked. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## NVIDIA H100 An NVIDIA graphics processing unit (GPU) built on the company's Hopper architecture and designed to accelerate AI workloads, including large language models. - [Definition reference](https://www.nvidia.com/en-us/data-center/h100/) ## Neocloud A new breed of cloud compute provider focused on renting out GPU computing power: a pure-play GPU cloud whose economics are still evolving. - [Definition reference](https://newsletter.semianalysis.com/p/ai-neocloud-playbook-and-anatomy) ## Editorial provenance Version 1. Publication ID: fefc142a-0f77-8ff7-1467-fd1a9f12a9d5. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-06T13:08:49.229Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model. Observation: 07ab61ba2ab4d825c1e0483fe7f2136c7129a8903f82bb860ed0beeb5ab25c2e. Source text hash: 63961be855484ae3781213bda9fd5fa9b8e506eb79796946a20e05878e176a68. ## Revision and source-check history ### Version 1 — 2026-11-06T13:13:24.206Z Carmen Li’s article in Silicon Data was published June 16, 2026. She argues that H100 rental rates need separate Hyperscaler and Neocloud indices because the same chip’s two market tiers behaved differently over the April–May window. The Hyperscaler index moved +0.27% end to end, within a roughly 2% range; the Neocloud index moved +3.8%, with an 11.7% trough-to-peak swing. Li attributes the contrast to market structure: longer reservations and an administered price ceiling make the Hyperscaler rate slow to reprice, while shorter commitments and more price-sensitive buyers let Neocloud prices respond more quickly to supply and demand. She says the indices serve different purposes: the Hyperscaler series can anchor reservation pricing and budgeting, while the Neocloud series can inform month-to-month rental decisions and offer an earlier read on demand. Li’s conclusion is that blending the indices would average away the stability of one and the responsiveness of the other. She cautions that the series show price movements, not their cause. Whether Neocloud strength continues remains unknown; Li says June data may help answer that question. ## Original source Silicon Data · Article Original source date: Jun 16, 2026 (2026-06-16). Date basis: Date stated on the publisher’s page. Added to Escape Meridian: 2026-11-06T13:13:24.206Z. First observed: 2026-10-25T01:53:38.909Z. Retrieved: 2026-10-25T01:55:38.181Z. - [H100 Index, May 2026: Same Chip, Two Tiers, Two Regimes - Silicon Data — GPU Performance Data for Companies](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Limits Text as retrieved on October 25, 2026; originally published June 16, 2026. Li presents the structural explanations for the different price behavior as likely or plausible; the article says the index movement itself does not establish its cause. The article says June data may help answer whether Neocloud strength continues; that outcome is not established in this reading. ## Open question What will June data show about whether Neocloud strength continues? ## Source-stated Carmen Li argues that the two H100 rental-rate tiers behaved differently over the same April–May window. > Over those two months, the Hyperscaler index moved +0.27% end to end, inside a peak-to-trough range of about 2%. The Neocloud index moved +3.8% end to end, but over a far wider path: - [Supporting source](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Source-stated Li attributes the difference in price movement to distinct market structures, including longer reservations and a price ceiling for Hyperscalers versus shorter commitments in the Neocloud tier. > The Hyperscaler tier functions as a price ceiling that AWS, Azure, and Google Cloud post above a large book of already-priced reserved capacity. - [Supporting source](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Source-stated Li says the indices serve different uses: the Hyperscaler series as a budgeting reference and the Neocloud series for spot-exposed decisions and demand monitoring. > The Hyperscaler index suits structured reservation pricing and enterprise budgeting. Its stability is the point: - [Supporting source](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Source-stated Li argues that combining the two series would average away the distinct properties that make each useful. > Blending them into one number would average away the stability that makes the Hyperscaler series a budgeting anchor and the responsiveness that makes the Neocloud series an earlier read on demand. - [Supporting source](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Source-stated Li says the price series show the Neocloud move but do not establish its cause. > We note this as an interpretation; the series itself shows the move, not its cause. - [Supporting source](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Source-stated The article was written by Carmen Li and published June 16, 2026. > Written by Carmen Li > Founder at Silicon Data > \# Industry Jun 16, 2026 8 Mins Read - [Supporting source](https://www.silicondata.com/blog/h100-index-may-2026-two-tiers-two-regimes) ## Source and contributor context Silicon Data · Publisher The publisher is identified here. No individual author has been inferred. ## Source media Media retrieval status: unchecked. ## Terms and references Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement. ## NVIDIA H100 An NVIDIA graphics processing unit (GPU) built on the company's Hopper architecture and designed to accelerate AI workloads, including large language models. - [Definition reference](https://www.nvidia.com/en-us/data-center/h100/) ## Neocloud A new breed of cloud compute provider focused on renting out GPU computing power: a pure-play GPU cloud whose economics are still evolving. - [Definition reference](https://newsletter.semianalysis.com/p/ai-neocloud-playbook-and-anatomy) ## Editorial provenance Version 1. Publication ID: fefc142a-0f77-8ff7-1467-fd1a9f12a9d5. Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5. Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-06T13:08:49.229Z. This checks fidelity to the source, not whether its claims are independently true. Editorial selection: accept; reasons: bridge\_admitted\_by\_model. Observation: 07ab61ba2ab4d825c1e0483fe7f2136c7129a8903f82bb860ed0beeb5ab25c2e. Source text hash: 63961be855484ae3781213bda9fd5fa9b8e506eb79796946a20e05878e176a68. ### uncertain — 2026-11-06T13:08:49.229Z — anthropic/claude-opus-5.5 Publication: fefc142a-0f77-8ff7-1467-fd1a9f12a9d5 attribution — pass: The reading correctly attributes the article and its arguments to Carmen Li, founder of Silicon Data, publishing on Silicon Data's blog. It is a single article, not a listing. > Written by Carmen Li Founder at Silicon Data summary — uncertain: The summary is faithful and keeps the source's hedges, conclusion and open question. However, 'Hyperscaler' is a central term that is neither explained in the reading nor in the supplied definitions. A general reader may not know it means the big cloud providers (AWS, Azure, Google Cloud), which the source names. Smallest repair: identify the Hyperscaler tier as AWS, Azure and Google Cloud, or add a term-library definition. 'Administered price ceiling' would also benefit from a brief plain-language gloss. > The Hyperscaler index covers the enterprise-SLA tier (AWS, Azure, Google Cloud) and behaves like an administered price ceiling date — pass: originalPublished is 2026-06-16 at day precision, which matches the article dateline. The reading states the date at that precision and uses no hindsight framing. The June outcome is correctly left as an open question. The evidence is a publisher page, so no archive disclosure is needed. > \# Industry Jun 16, 2026 8 Mins Read support — pass: Every figure and claim matches the source: +0.27%, about 2% range, +3.8%, the 11.7% swing, the structural explanations, the separate uses, the blending argument and the cause caveat. The limitations correctly note that Li offers the structural explanations as likely or plausible. The cause caveat applies specifically to the Neocloud move, and the claims preserve this. > Over those two months, the Hyperscaler index moved +0.27% end to end, inside a peak-to-trough range of about 2%. media — pass: No media was selected, and no verified candidates were supplied, so nothing is misidentified. ## Recorded references Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter. No public references recorded here yet. [Reference data (JSON)](/readings/c317825d1d3b8363fdabb1427f562af06a8893e7da349092e98acae811d09fbc/references.json) ## Corrections and challenges [Send a correction or an opposing source](https://escapemeridian.com/corrections/) [JSON record](https://escapemeridian.com/readings/carmen-li-argues-in-silicon-data-for-separate-h100-rental-rate-c317825d.json)