> ## Content Index > Fetch the complete content index at: https://www.sotkn.com/llms.txt > Use this file to discover other available public pages before exploring further. # 4.8× Tokens, 1.5 GW Live: Who Gets Paid When US Halls Still Bind - URL: https://www.sotkn.com/markets/us/us-coreweave-vera-rubin-october-2026/ - Published: 2026-10-01T02:11:01.000Z - Updated: 2026-10-01T02:12:31.000Z - Description: October 2026 in one line: CoreWeave put Vera Rubin NVL72 in production. Who gets paid is the live 1.5 GW. The stuck step is still contracted halls. - Author: Sean - Tags: #market-us CoreWeave put NVIDIA Vera Rubin NVL72 into a paying customer's hall. Here is who gets paid, which step is still stuck, and the arithmetic before the next megawatt print. US PULSE · Data as at 1 Oct 2026 · Next update: after CoreWeave Q3 results (date TBC) Affects: Neoclouds · Hyperscalers · Semiconductors — Markets: US · Taiwan · Korea **In short:** On 30 September 2026, [CoreWeave](https://www.coreweave.com/news/coreweave-delivers-nvidia-vera-rubin-nvl72-performance-at-production-scale-starting-with-cognition?ref=sotkn.com) said NVIDIA Vera Rubin NVL72 is available on its cloud, with Cognition the first customer anywhere running production workloads. Cognition measured up to **4.8 times** total token throughput on SWE-2 inference versus a GB200 NVL72 baseline, and **3.8 times** output-token throughput on reinforcement learning. Who gets paid is NVIDIA, CoreWeave, Cognition, and the landlords already collecting on energised megawatts. The stuck step is still **contracted halls that are not active** — 1.5 GW live against about 3.7 GW contracted as of 30 June 2026\. The next hard clocks are CoreWeave's Q3 megawatt print and the 16 November RTO large-load filings. *Education only, not advice to buy or sell any security. Figures are from company and government releases dated as shown.* ## The numbers that matter this month | Print | Figure | Read | | ------------------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | Dated trigger | CoreWeave, 30 Sep 2026 | Vera Rubin NVL72 in production; Cognition first customer | | Cognition SWE-2 inference | 4.8× tokens vs GB200 NVL72 | Customer-run benchmark on CoreWeave Cloud | | Cognition RL | 3.8× output tokens vs GB200 | Same print; different workload | | CoreWeave silicon claim | 10× tokens per MW vs GB200 | DeepSeek R1, matched interactivity · company, not Cognition | | Active power (30 Jun) | 1.5 GW | [Q2 print, 11 Aug 2026](https://investors.coreweave.com/news/news-details/2026/CoreWeave-Reports-Strong-Second-Quarter-2026-Results/default.aspx?ref=sotkn.com) | | Contracted power (30 Jun) | \~3.7 GW | Same Q2 print | | Unsatisfied RPO (30 Jun) | $103.7bn | 10-Q: 41% in 24 months | | Helios Phase I | 133 MW IT / 200 MW gross | [Galaxy, 5 Aug 2026](https://www.sec.gov/Archives/edgar/data/1859392/000185939226000084/glxy-20260630xpressrelease.htm?ref=sotkn.com) · $80m quarterly rent guided | | Helios CoreWeave lease | 526 MW IT / 800 MW gross | Phases I+II+III · 15-year base | | Same-week official twin | BEA Q2 GDP third estimate | 2.2%; structures revision led by data centres | | Separate US clock | Micron FQ4 | $54.23bn revenue · do not mix into halls | Status chips are Second Order's read. Full sources are at the end of this article. ## Why does the US matter now? Because the US is where the newest racks are supposed to light — and this week a named customer ran production on Vera Rubin while most of the contracted US power book was still waiting for a hall. Taiwan still makes most of the leading chips and does much of the [CoWoS packaging](https://www.sotkn.com/explained/cowos-explained/). Korea still allocates [HBM](https://www.sotkn.com/explained/hbm-explained/). The US is where those chips convert into billed hours. When CoreWeave can stand up a Rubin cluster in early September and have Cognition benchmarking it against GB200 within days, the US door is not "more GPUs." It is which megawatts are already on. The buyers are the [hyperscalers](https://www.sotkn.com/hyperscalers/) and the [neoclouds](https://www.sotkn.com/neoclouds/). NVIDIA ships the rack. CoreWeave sells the hour. Galaxy collects rent on the 133 MW that is actually in service. If you only remember one thing: **a faster rack does not unstick a dark campus.** New to the US clock? [Speed-to-power, explained](https://www.sotkn.com/explained/speed-to-power-explained-for-investors/). Siblings: [US Speed-to-Power](https://www.sotkn.com/markets/us/us-speed-to-power-september-2026/) · [AI labs, small MW](https://www.sotkn.com/markets/us/us-ai-labs-small-mw-september-2026/) · [Joliet credit](https://www.sotkn.com/markets/us/us-ferc-joliet-credit-september-2026/). ## What changed The dated print is CoreWeave's 30 September 2026 release: NVIDIA Vera Rubin NVL72 available, Cognition first in production. That is not a campus announcement. It is a production claim on a named workload. From the release: Cognition, the lab behind Devin, stood up a Vera Rubin NVL72 cluster on CoreWeave in early September. Its engineers ran the first customer-executed Vera Rubin inference benchmark against a GB200 NVL72 baseline. SWE-2 inference printed 4.8 times total token throughput. Reinforcement-learning workloads printed 3.8 times output-token throughput. CoreWeave also restated its own silicon number: 10 times token throughput per megawatt versus GB200 NVL72 on DeepSeek R1 at matched interactivity. Treat that as a company measurement on a different model. Do not staple it to Cognition's 4.8 times. What did **not** change on 30 September is the power book. The last hard megawatt print is still 11 August: active power 1.5 GW, contracted power about 3.7 GW, unsatisfied remaining performance obligations $103.7 billion as of 30 June. Galaxy's 5 August print still has Helios Phase I at 133 MW of critical IT load — 200 MW gross — with Phase II's 260 MW of IT not due to start delivering until the second quarter of 2027. The same morning, [BEA](https://www.bea.gov/news/2026/gdp-third-estimate-industries-corporate-profits-state-gdp-and-state-personal-income-2nd?ref=sotkn.com) raised Q2 2026 real GDP to 2.2% in the third estimate. The structures revision was "led by commercial and health care, mainly data centers," on revised Census construction put-in-place for May and June. That is the official twin, not the trigger: halls being poured is not halls energised. [Micron](https://investors.micron.com/news/press-release/2026/Micron-Technology-Inc--Reports-Record-Fiscal-Fourth-Quarter-and-Full-Year-2026-Results/default.aspx?ref=sotkn.com) reported fiscal Q4 the same day: $54.23 billion of revenue, FQ1-27 guidance $61.5 billion ± $1.5 billion. That is a US memory clock. It is not this Pulse. HBM still sits on the Korea door. ## Our math Four calculations — labelled **our arithmetic** — put the dated prints on one unit. **1\. How much of the contracted book is actually on.** 1.5 ÷ 3.7 = **40.5%**. 3.7 − 1.5 = **2.2 GW** contracted, not yet active. That is the conversion gap. Q2 said active power rose by nearly 500 MW to reach 1.5 GW. The contracted book is still more than twice the live book. A production Rubin cluster is evidence that the live slice can take a new generation. It is not evidence that 2.2 GW moved. **2\. Helios as a named campus inside that gap.** Galaxy's CoreWeave lease across Phases I+II+III is 526 MW of critical IT (800 MW gross) on a 15-year base, against 1.63 GW of approved gross campus power. 133 ÷ 526 = **25.3%** of the leased IT load is in service. 133 ÷ 1,500 = **8.9%** of CoreWeave's active book is this one Phase I. 526 − 133 = **393 MW** of leased IT still to deliver. CoreWeave's Q2 10-Q described 393 MW of electrical power remaining undelivered at a single site, expected in phases in 2026 and 2028\. The numbers match. Treat the match as our read, not as Galaxy restating the 10-Q. Phase II is 260 MW of critical IT, with hall deliveries expected to begin in Q2 2027\. That is a next-year clock, not a 30 September clock. **3\. Dollars per live megawatt — landlord versus neocloud.** Galaxy guided about $80 million of quarterly Phase I leasing revenue on 133 MW. $80,000,000 ÷ 133 = **$602,000 per MW per quarter** for the landlord. CoreWeave printed $2.575 billion of Q2 revenue on 1.5 GW of active power. $2,575,000,000 ÷ 1,500 = **$1.72 million per MW per quarter** for the neocloud. $103.7 billion of unsatisfied RPO ÷ 1,500 MW = **$69.1 million of backlog per active MW**. $103.7 billion ÷ 3,700 MW = **$28.0 million per contracted MW**. The 10-Q says 41% of that RPO is expected in the 24 months to 30 June 2028: 0.41 × $103.7 billion = **$42.5 billion** in the near window — our arithmetic, still subject to delivery. The live megawatt is the scarce object. [Take-or-pay](https://www.sotkn.com/explained/take-or-pay-gpu-contracts-explained/) and [RPO](https://www.sotkn.com/explained/rpo-vs-take-or-pay/) only convert when the seat is powered. That is the [CoreWeave contracts Pulse](https://www.sotkn.com/markets/us/coreweave-neocloud-contracts/). **4\. Tokens per megawatt as a bypass, not a substitute.** CoreWeave's 10× tokens-per-MW claim versus GB200, if taken at face value on that DeepSeek R1 test: 133 × 10 = **1,330 MW-equivalent** of GB200-class tokens from Helios Phase I's 133 MW of IT. Cognition's 4.8× on SWE-2, applied to a 30 MW speed-lane block: 30 × 4.8 = **144 MW-equivalent** of GB200 SWE-2 work from one small-MW hall. Do not add those products. They are different models and neither is a utility interconnection. The so-what is narrower: on a hall that already has power, a new rack raises tokens without waiting for the next transformer. That is why labs still hunt 20–30 MW at powered sites — the same bypass as the [small-MW Pulse](https://www.sotkn.com/markets/us/us-ai-labs-small-mw-september-2026/), now with a named generation on it. What this does not tell you: the next active-power print. We do not do price calls and we do not forecast a Q3 slide. ![Four-card chart of Second Order arithmetic: 40.5% active share, 2.2 GW waiting, $69.1m RPO per live MW, 25.3% of Helios lease in service](https://storage.ghost.io/c/cd/10/cd10e222-131f-4ba4-b4a9-ca4214e8c12d/content/images/2026/10/us-coreweave-vera-rubin-october-2026-math.png) *If the Q2 power book still holds, our arithmetic is 1.5 of 3.7 GW live. Sources: CoreWeave 11 Aug 2026; Galaxy 5 Aug 2026\. Education only.* ![Bar chart comparing 30 MW small deals, 133 MW Helios Phase I, 1.5 GW active, 1.8 GW Joliet, 3.7 GW contracted](https://storage.ghost.io/c/cd/10/cd10e222-131f-4ba4-b4a9-ca4214e8c12d/content/images/2026/10/us-coreweave-vera-rubin-october-2026-scale.png) *A production Rubin cluster sits inside a 2.2 GW contracted wait. Sources: CoreWeave; Galaxy; FERC; CNBC. Education only.* ## Where does the money stick? At the steps with the fewest good substitutes — and, this month, at the megawatts that are already on. ![Infographic: live Rubin racks get paid, contracted dark halls are the stuck step, small MW still clears now](https://storage.ghost.io/c/cd/10/cd10e222-131f-4ba4-b4a9-ca4214e8c12d/content/images/2026/10/us-coreweave-vera-rubin-october-2026-infographic.png) *Scarce usable megawatts keep more of each dollar. A 4.8× rack does not energise 2.2 GW.* **1\. Live racks — who gets paid now.** NVIDIA gets paid on Vera Rubin silicon that a customer will actually run. CoreWeave gets paid on the hour once the cluster is in a hall. Cognition gets more Devin sessions per GPU on the SWE-2 print. Dell gets paid on PowerRack. Galaxy gets paid on the 133 MW already in service — about $80 million a quarter on Phase I, on its own guide. Those are cash steps. They do not require the 2.2 GW to move. **2\. Contracted, not-yet-active halls — the stuck step.** 3.7 GW contracted minus 1.5 GW active is the US conversion problem in one line. Helios makes it concrete: 25% of the leased IT is on; Phase II does not start delivering until 2027; 830 MW of approved Helios capacity is not even under lease. The 10-Q's 393 MW undelivered at one site is the same object. Announced gigawatts and contracted gigawatts are not billed hours. **3\. Credit paper still sits in front of some campuses.** Last week's [Joliet Pulse](https://www.sotkn.com/markets/us/us-ferc-joliet-credit-september-2026/) is not retired: 1.8 GW of Illinois paper is still a docket. A Rubin bring-up in a live hall does not study that load. **4\. Small MW and powered shells.** The 18 September CNBC print — labs hunting 20–30 MW — is the speed lane: more tokens in a hall you already have, while GW leases wait for 2027 steel. The pattern is the same as Taiwan and Korea, on a different object: **the money sticks where the step is stuck.** This week the stuck step is the hall that is contracted and not on. ## Is contracted-but-not-active power still stuck? Yes. The 30 September print is evidence that the live slice can take a new GPU generation. It is not evidence that the conversion gap closed. Q2 already said CoreWeave had completed the industry's first Vera Rubin NVL72 bring-up. 30 September adds a named production customer and two customer-run multiples. The power clock is still 1.5 / 3.7, the 393 MW undelivered, and Galaxy's Q2 2027 Phase II start. Two cautions before you read too much into "Rubin is in production": - **A cluster is not a campus.** Cognition's early-September stand-up is a live room. Helios's remaining 393 MW of leased IT is not. - **Tokens per MW are not interconnection.** 10× on DeepSeek R1, or 4.8× on SWE-2, raises output on power you already have. It does not create a substation. If contracted halls catch up — active power closing on 3.7 GW, Helios Phase II pulling forward — the stuck step moves. Next candidates: transformers, credit-backed interconnection, then memory. Packaging remains a Taiwan door. ## What dates matter The next clocks are a megawatt print, a memory print we will not mix in, and the same tariff clock as Joliet. ![Calendar of 30 Sep CoreWeave print, Micron FQ4, 8 Oct TSMC, CoreWeave Q3 TBC, 16 Nov RTO filings](https://storage.ghost.io/c/cd/10/cd10e222-131f-4ba4-b4a9-ca4214e8c12d/content/images/2026/10/us-coreweave-vera-rubin-october-2026-calendar.png) *The rack is live. Active-versus-contracted power still sets the clock.* | Date | What | Why it matters | | ----------- | ---------------------------------- | --------------------------------------------------------------------------------------------- | | 30 Sep 2026 | CoreWeave Vera Rubin in production | This Pulse's print. Cognition 4.8× / 3.8×. | | 30 Sep 2026 | Micron fiscal Q4 | $54.23bn. US memory — separate clock. | | 30 Sep 2026 | BEA Q2 GDP third estimate | 2.2%; data-centre structures in the revision. Official twin. | | 8 Oct 2026 | TSMC September revenue | Taiwan door — packaging still sets who ships. | | Open / TBC | CoreWeave Q3 2026 results | Active MW, contracted MW, RPO. Street dates cluster in November; CoreWeave has not confirmed. | | 16 Nov 2026 | RTO / ISO large-load filings | Pro forma credit and cost-recovery for the next decade of GW campuses. | | Q2 2027 | Helios Phase II hall deliveries | Galaxy: 260 MW IT, start of delivery. | ## How is this different from last week's US Pulses — and from Taiwan and Korea? It is the same map. It is a different object. The [19 September Speed-to-Power Pulse](https://www.sotkn.com/markets/us/us-speed-to-power-september-2026/) mapped the physical clock — transformers, ERCOT's queue, contracted power running ahead of energised power. This Pulse puts a named GPU generation on the live side of that split. The [21 September small-MW Pulse](https://www.sotkn.com/markets/us/us-ai-labs-small-mw-september-2026/) mapped the bypass: labs hunting 20–30 MW at powered sites. Cognition's 4.8× is what that bypass is for. The [25 September Joliet Pulse](https://www.sotkn.com/markets/us/us-ferc-joliet-credit-september-2026/) mapped the credit clause in front of 1.8 GW of paper — larger than CoreWeave's entire 1.5 GW of active power. A production Rubin cluster and a $1 letter of credit are not the same product. Taiwan's door is still [CoWoS](https://www.sotkn.com/markets/taiwan/taiwan-tsmc-revenue-september-2026/). Korea's door is still [HBM allocation](https://www.sotkn.com/markets/korea/korea-chip-output-august-2026/). Micron's $54 billion quarter is a US memory print sitting next to that door, not inside this Pulse. [CoreWeave backlog → cash](https://www.sotkn.com/markets/us/coreweave-neocloud-contracts/) is the sister sentence: contracts become cash when halls are power-ready. 30 September says the live halls can take Rubin. It does not say the backlog's other 2.2 GW is ready. ## What would change this view? The "contracted-but-not-active halls are the US stuck step" read weakens if: - \[ \] CoreWeave's next megawatt print shows active power closing most of the 2.2 GW gap, not just another 100–200 MW - \[ \] Galaxy pulls Helios Phase II deliveries into 2026, or a second named campus prints 200 MW+ of new critical IT in service - \[ \] Labs stop paying up for 20–30 MW powered blocks because GW halls start lighting this year - \[ \] The Northern District of Illinois and the mid-November RTO filings make large-load paper boring, *and* transformers stop being the louder constraint - \[ \] A non-power bottleneck — HBM, CoWoS, or cooling — becomes the louder US constraint on who gets paid this quarter - \[ \] Cognition-class customers stop taking new NVIDIA generations on neoclouds and wait for hyperscaler SKUs instead The view stays intact while 1.5 GW sits under 3.7 GW contracted, Helios Phase II is 2027, Joliet is still a docket, and a 4.8× rack is what you do in a room you already have. ## Practical takeaways - Name the stuck step: **contracted halls that are not active** — not "the US ran out of GPUs" - Who gets paid: NVIDIA, CoreWeave, Cognition, Dell, and landlords already collecting on live MW (Galaxy's $80m quarterly Phase I guide) - Who waits: 2.2 GW of contracted CoreWeave power, 393 MW of Helios leased IT, and 1.8 GW of Joliet paper - Our arithmetic: 1.5 ÷ 3.7 = 40.5%; 133 ÷ 526 = 25.3%; $103.7bn ÷ 1,500 MW = $69.1m of RPO per live MW; 30 × 4.8 = 144 MW-equivalent of GB200 SWE-2 work in a small hall - Read beside [Speed-to-Power](https://www.sotkn.com/markets/us/us-speed-to-power-september-2026/), [small MW](https://www.sotkn.com/markets/us/us-ai-labs-small-mw-september-2026/), and [Joliet](https://www.sotkn.com/markets/us/us-ferc-joliet-credit-september-2026/) — same conversion problem, four lenses - CoreWeave Q3 (date TBC) is the next megawatt clock; 16 November is the next industry-wide paper clock ## FAQ **What did CoreWeave announce on 30 September 2026?** NVIDIA Vera Rubin NVL72 is available on CoreWeave, with Cognition the first customer running production workloads. Cognition measured 4.8× total token throughput on SWE-2 inference and 3.8× output-token throughput on RL versus GB200 NVL72. **Who gets paid on this US print?** NVIDIA on the rack, CoreWeave on the billed hour, Cognition on more sessions per GPU, Dell on PowerRack, and landlords already collecting on energised megawatts (Galaxy guided \~$80 million a quarter on Helios Phase I). A 2.2 GW contracted wait does not get paid by a customer benchmark. **Why are halls still the stuck step if Rubin is in production?** Because production here means a named cluster in a hall that already had power. As of 30 June, CoreWeave had 1.5 GW active against about 3.7 GW contracted. Helios Phase I is 133 MW of 526 MW leased IT. The rest is still a delivery clock. **What is the difference between active power and contracted power?** Active power is the megawatts CoreWeave says are on. Contracted power is the megawatts it has the right to take. RPO and take-or-pay sit on top of that split: cash shows up when the seat is powered and the hour is sold. **How is this different from the US Speed-to-Power Pulse?** Speed-to-Power maps the physical clock — transformers, queues, energised halls. This Pulse maps a named GPU generation onto the live side of that clock, and keeps the 2.2 GW contracted gap as the stuck step. **Is this a buy list of CoreWeave, NVIDIA, or data-centre stocks?** No. This Pulse maps who gets paid and which step is stuck. It is education only, not a recommendation to buy or sell any security. ## Keep reading on the map - **Explained:** [Speed-to-power](https://www.sotkn.com/explained/speed-to-power-explained-for-investors/) · [Take-or-pay GPU contracts](https://www.sotkn.com/explained/take-or-pay-gpu-contracts-explained/) · [RPO vs take-or-pay](https://www.sotkn.com/explained/rpo-vs-take-or-pay/) · [HBM, explained](https://www.sotkn.com/explained/hbm-explained/) · [CoWoS, explained](https://www.sotkn.com/explained/cowos-explained/) - **Markets:** [US Speed-to-Power](https://www.sotkn.com/markets/us/us-speed-to-power-september-2026/) · [US small MW](https://www.sotkn.com/markets/us/us-ai-labs-small-mw-september-2026/) · [US Joliet credit](https://www.sotkn.com/markets/us/us-ferc-joliet-credit-september-2026/) · [CoreWeave contracts](https://www.sotkn.com/markets/us/coreweave-neocloud-contracts/) · [Korea chip output](https://www.sotkn.com/markets/korea/korea-chip-output-august-2026/) · [Taiwan TSMC revenue](https://www.sotkn.com/markets/taiwan/taiwan-tsmc-revenue-september-2026/) - **The Stack:** [Hyperscalers](https://www.sotkn.com/hyperscalers/) · [Neoclouds](https://www.sotkn.com/neoclouds/) · [Semiconductors](https://www.sotkn.com/semiconductors/) **Members' Analysis.** If new GPU generations keep lighting in halls that already have power while 2.2 GW waits on 2027 steel, the bottleneck is usable megawatts — not silicon. [Read the Analysis →](https://www.sotkn.com/analysis/) **Get the Wire.** [Subscribe free →](https://www.sotkn.com/#/portal/signup) ### How we check this Figures come from company filings and government releases, each dated in the list below. Capacity and timing from press reports are labelled as reported, not guided. Where we work something out ourselves, the arithmetic is shown in full. **Last reviewed:** 1 Oct 2026 · **Next review:** after CoreWeave Q3 results or the next named megawatt delivery print Spot an error? [Tell us](mailto:sean@thekopinotes.com) and we will correct it and note the change here. [Our editorial standards →](https://www.sotkn.com/editorial-standards/) ### Sources - [CoreWeave — Vera Rubin NVL72 in production with Cognition](https://www.coreweave.com/news/coreweave-delivers-nvidia-vera-rubin-nvl72-performance-at-production-scale-starting-with-cognition?ref=sotkn.com) (30 Sep 2026) - [CoreWeave — Q2 2026 results](https://investors.coreweave.com/news/news-details/2026/CoreWeave-Reports-Strong-Second-Quarter-2026-Results/default.aspx?ref=sotkn.com) (11 Aug 2026; 1.5 GW active, \~3.7 GW contracted, \~$104bn backlog) - [CoreWeave — Q2 2026 earnings 8-K exhibit](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000362/coreweave2q26earningspress.htm?ref=sotkn.com) - [CoreWeave Form 10-Q](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000366/?ref=sotkn.com) for the quarter ended 30 Jun 2026 (unsatisfied RPO $103.7bn, 41/39% windows; 393 MW undelivered at one site) - [Galaxy Digital — Q2 2026 results / Helios](https://www.sec.gov/Archives/edgar/data/1859392/000185939226000084/glxy-20260630xpressrelease.htm?ref=sotkn.com) (5 Aug 2026; 133 MW IT Phase I, 526 MW leased IT, 1.63 GW approved) - [BEA — GDP third estimate, Q2 2026](https://www.bea.gov/news/2026/gdp-third-estimate-industries-corporate-profits-state-gdp-and-state-personal-income-2nd?ref=sotkn.com) (30 Sep 2026) - [Micron — fiscal Q4 and FY 2026 results](https://investors.micron.com/news/press-release/2026/Micron-Technology-Inc--Reports-Record-Fiscal-Fourth-Quarter-and-Full-Year-2026-Results/default.aspx?ref=sotkn.com) (30 Sep 2026) - [CNBC — Anthropic / OpenAI 20–30 MW hunt](https://www.cnbc.com/2026/09/18/anthropic-openai-small-ai-data-center-deals.html?ref=sotkn.com) (18 Sep 2026) - [FERC — large-load show-cause orders](https://www.ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration?ref=sotkn.com) (18 Jun 2026) - Second Order — [US Speed-to-Power](https://www.sotkn.com/markets/us/us-speed-to-power-september-2026/) · [US small MW](https://www.sotkn.com/markets/us/us-ai-labs-small-mw-september-2026/) · [US Joliet](https://www.sotkn.com/markets/us/us-ferc-joliet-credit-september-2026/) · [Korea chip output](https://www.sotkn.com/markets/korea/korea-chip-output-august-2026/) Education only. Not investment advice. Do your own research. Second Order may correct figures if primary sources revise them.