How many systems on the November 2026 TOP500 list — Metaculus Scans

Plan 6.78s

Sub-questions (0)

Tool requests (3)

resolution_source {"question": "How many systems on the November 2026 TOP500 list will be exascale?", "source_name": "TOP500 list (June 2026 edition as current status quo)", "urls": ["https://www.top500.org/lists/top500/", "https://www.top500.org/news/lineshine-debuts-no-1-top500-enters-new-global-exascale-era/"]}

wikipedia_lookup {"queries": ["TOP500", "Exascale computing", "JUPITER supercomputer", "Alice Recoque supercomputer"]}

claude_news {"brief": "Identify supercomputers expected to debut or be upgraded with HPL Rmax \u003e= 1 EFlop/s on the November 2026 TOP500 list (SC26). Candidates include Alice Recoque, Discovery/Doudna, Horizon, hyperscaler AI clusters, Japanese and other new systems, and further runs of JUPITER, Eagle and LineShine. List each system\u0027s reported or expected Rmax, and say which systems sit just below 1 EF and could cross it.", "max_searches": 4, "question_title": "How many systems on the November 2026 TOP500 list will be exascale?"}

Research 50.44s · 3 tool(s) · 0 error(s)

resolution_source ok resolved=unclear; The November 2026 TOP500 list has not been publish 10.06s

wikipedia_lookup ok 3 pages 2.11s

claude_news ok 13 citations 32.33s

research loops 6/6 ok · 27 searches · 9 page reads

current state ok 5 searches · 1 reads · 3884 chars 20.2s

base rate ok 5 searches · 1 reads · 4276 chars 21.7s

key factors ok 4 searches · 2 reads · 4565 chars 20.1s

expert view ok 5 searches · 1 reads · 4201 chars 21.0s

favorite case ok 4 searches · 2 reads · 4329 chars 24.6s

upset case ok 4 searches · 2 reads · 4771 chars 20.8s

report: current state

- [2026-06-23] The June 2026 list (67th edition) is the latest published TOP500 list, and it has five systems at Rmax ≥ 1 EF/s. They are LineShine 2.198, El Capitan 1.809, Frontier 1.353, Aurora 1.012 and JUPITER Booster 1.000 EF/s. TOP500 says LineShine raised the count from four to five. https://www.top500.org/news/lineshine-debuts-no-1-top500-enters-new-global-exascale-era/
- [2026-06-23] JUPITER Booster sits exactly at the threshold. TOP500 lists it at 1,000.00 PFlop/s Rmax, so it meets the "≥1,000" criterion with no margin. https://www.top500.org/lists/top500/2026/06/highs/
- [2026-06-23] The system at No. 6 is far below the threshold: Eni HPC7 at 571.5 PF/s. Eagle is at 561.2 and HPC6 at 477.9 PF/s. No other system is within about 40% of exascale, so a new entrant from this tier would need an unannounced jump. https://www.top500.org/news/lineshine-debuts-no-1-top500-enters-new-global-exascale-era/
- [2025-11-17] The previous list (November 2025) had four exascale systems. JUPITER crossed 1 EF/s then, up from a 793 PF/s partial run in June 2025. The count has gone 3 (June 2025), 4 (Nov 2025), 5 (June 2026). https://www.theregister.com/2025/11/17/europe_jupiter_supercomputer/
- [2026-06-23] The only other publicly named exascale-class build that I found is Alice Recoque (France, Eviden/AMD). Installation starts in 2026, and capacity testing and first users are expected around end-2027. It is unlikely to be on the November 2026 list. https://www.surf.nl/en/themes/research-infrastructure/alice-recoque
- [2026-06-25] JUPITER's technical page says installation of its components is expected to conclude in 2026. The sources I found do not show a new full-system HPL run. https://www.fz-juelich.de/en/jsc/jupiter/tech
- [2026-10-12] Fontanka reports that drone attacks on Yandex data centres (Oct 8–11) may have damaged three Russian systems. The article gives them as Chervonenkis at No. 101, Galushkin at No. 134 and Lyapunov at No. 161. All are at or below roughly 21 PF/s, so none is near exascale. Yandex has not confirmed the damage. https://www.fontanka.ru/2026/10/12/76696055/
- [2026-10-13] The TOP500 homepage still shows the June 2026 list as current, with five exascale systems. No November 2026 list or preview has been published. https://top500.org/
- [2026-11-15] SC26 runs Nov 15–20, 2026 at McCormick Place, Chicago, with exhibits Nov 17–19. TOP500 lists are normally released at the SC opening, so Nov 16 is the likely release date, but I did not find an official date. https://sc26.supercomputing.org/attendees/chicago/
- [2026-09-14] Chinese submissions are voluntary. Older reports of Sunway OceanLight (~1.05 EF/s) and Tianhe-3 (~1.3 EF/s) suggest China holds exascale systems that are not on the list. LineShine was China's first submission in about nine years. A further Chinese submission is possible but unsignalled. I found no 2026 reporting on it. https://chipsandcheese.com/p/top500-at-isc26-we-have-a-new-number
- [2026-08-06] A DataCenterKnowledge piece says the next list comes at SC26 and is "the next opportunity" for a challenger to LineShine. It names no specific new exascale entrant. https://www.datacenterknowledge.com/supercomputers/flops-vs-megawatts-who-s-winning-in-2026-supercomputing-

BOTTOM LINE: The published baseline is 5 systems at Rmax ≥ 1 EF/s. No identified candidate is ready to add a sixth, and the main downside risk is JUPITER at exactly 1.000 EF/s or an existing system not re-running. Any change would have to come from an unannounced submission, for example a Chinese or hyperscaler system, or from a larger re-run of an existing machine.

NOT FOUND: any announced November 2026 submission or leak of a new ≥1 EF/s HPL run; an official release date for the 68th list; confirmation of whether JUPITER, Aurora and Frontier will re-run HPL; and anything newer than October 13 on the resolution source.

report: base rate

- [2026-06-23] The June 2026 TOP500 (67th list) had five systems with Rmax ≥ 1 EFlop/s, up from four. These were LineShine (2.198), El Capitan (1.809), Frontier (1.353), Aurora (1.012) and JUPITER Booster (1.000). HPC7, the next system, was at 0.5715. https://www.top500.org/news/lineshine-debuts-no-1-top500-enters-new-global-exascale-era/
- [2026-06-23] Margin on the threshold: JUPITER Booster sits at exactly 1,000.00 PFlop/s and Aurora at 1,012. The question says "≥1,000", so JUPITER counts. Both reappeared unchanged from the previous list. https://www.top500.org/lists/top500/2026/06/highs/
- [2026-06-23] The list the question cites counted five, and HPE also calls them "5 verified exascale systems". Of these, 3 are US (all DOE), 1 is in Germany and 1 is in China. https://www.hpe.com/us/en/newsroom/blog-post/2026/06/hpe-delivers-six-out-of-ten-of-worlds-most-powerful-supercomputers.html
- Historical count by list, from the TOP500 pages above and my own recollection (checked only for 2025–26): 0 until Jun 2022; 1 in Jun 2022–Nov 2023 (Frontier); 2 in Jun 2024 (+Aurora); 3 in Nov 2024 and Jun 2025 (+El Capitan). Then 4 in Nov 2025 (+JUPITER) and 5 in Jun 2026 (+LineShine). The count has grown by 0–1 per list and has never fallen. https://www.top500.org/lists/top500/2025/06/
- [2025-11] JUPITER entered at exactly 1.000 EFlop/s as the fourth exascale system. It was the first outside the US. https://www.techpowerup.com/343057/top500-el-capitan-retains-1-as-jupiter-becomes-europes-first-exascale-system
- [2026-06-25] LineShine was a surprise debut. China had not submitted to the TOP500 for about 9 years, and Chips and Cheese had not expected a new No. 1 before Discovery or Fugaku-Next. This suggests a further unannounced Chinese entry is possible, but I found no reporting on one. https://chipsandcheese.com/p/top500-at-isc26-we-have-a-new-number
- [2026-03 approx.] The community widely believes China has several exascale-class machines it does not submit. Treat this as an upside tail risk, not a forecast. https://metebalci.com/blog/state-of-supercomputers-around-the-world-in-2026/
- [2025-11-06] Alice Recoque (France/EuroHPC) is the main European candidate for a sixth system. Its own schedule has delivery starting Q4 2026, reaching 1 EFlop/s Linpack in Q2 2027, and operations in H2 2027. It is therefore very unlikely to be on the Nov 2026 list. https://numpex.org/wp-content/uploads/sites/6/2025/12/2025-11-06-02-Alice-Recoque.pdf
- [2026-06-23] The next-largest listed system is HPC7 at 571.5 PFlop/s, then Eagle at 561 PFlop/s. Reaching 1 EFlop/s would need roughly a 75% gain, which is unlikely in a re-measurement. El Capitan has already been re-measured upward once, from 1.742 to 1.809. https://www.top500.org/lists/top500/2026/06/
- [2026-10-13] The Nov 2026 list is due at SC26 in Chicago (Nov 15–20, with exhibits Nov 17–19). The question resolves on 2026-11-16, so I assume the list is released around then. Release timing is a small resolution risk. https://www.mathworks.com/company/events/tradeshows/international-conference-for-high-performance-computing-sc26-5180476.html
- [2026-10-12] Drone attacks hit Yandex data centers on Oct 8–11. Fontanka reports that three Russian A100 systems may be damaged, ranked #101, #134 and #161 on the June list. These systems are far below exascale, so this does not affect the count. https://www.fontanka.ru/2026/10/12/76696055/
- Downside risk is small. A drop below 5 would need a top-5 system to be withdrawn or to fall below its June Rmax. JUPITER at exactly 1.000 and Aurora at 1.012 have the thinnest margins. Neither dropped in earlier re-submissions.

BOTTOM LINE: The base rate points to 5 as the modal outcome, since the count has never fallen and has never risen by more than 1 per list. 6 would need an unannounced entry, most plausibly another Chinese system or an unexpectedly early US or EU installation. 4 is a thin tail, since JUPITER sits at exactly the threshold.

NOT FOUND: Any announced new ≥1 EFlop/s system scheduled for Nov 2026 (US Doudna, Discovery, Alice Recoque and Fugaku-Next all appear to be 2027 or later). Any expert or odds forecast of the Nov 2026 exascale count. Any report of a withdrawn or downgraded top-5 system. I did not confirm Doudna's timing.

report: key factors

- [2026-03-17] Doudna (NERSC-10, Dell/NVIDIA Vera Rubin) is the main US candidate for a sixth system. Berkeley Lab's own page says full-system deployment is "late 2026", with Doudna arriving "at the end of the year". The pilot early-access system, Cech, is only 5.76 PF FP64, delivered in January and still being installed. A full-scale Linpack run by the Nov 2026 submission deadline looks unlikely, and no source says one is planned. https://newscenter.lbl.gov/2026/03/17/berkeley-lab-takes-major-step-toward-doudna-with-delivery-of-early-access-system-cech/
- [2025-06-25] The Next Platform's estimate for Doudna is about 2.7 EF peak, inferred from a 20 MW envelope. That is plausibly exascale on HPL, but only once fully installed and tuned. Its timeline was the fall of 2026 for arrival at LBNL, then production in 2027. That makes Doudna a candidate for the Jun 2027 list rather than Nov 2026. https://www.nextplatform.com/2025/06/25/some-thoughts-on-the-future-doudna-nersc-10-supercomputer/
- [2025-11-18] TACC's Horizon (Blackwell/Vera) was billed as "online in 2026". It has 4,000 Blackwell GPUs. It was earlier described as a "half-exaflop" system, so it is unlikely to reach 1 EF Rmax on HPL. https://www.storagenewsletter.com/2025/11/18/sc25-nvidia-accelerates-ai-for-over-80-new-science-systems-worldwide/
- [2026-06-23] On the fetched TOP500 release, the table of exascale systems matches the five in the earlier reports. It also says LineShine was a "previously unlisted" system. That is the mechanism by which a new Chinese entry would arrive: with no pre-announcement and no pipeline. The release does not cite any Chinese pipeline. The same page says JUPITER Booster is at "exactly 1.000" and still has no HPCG result. https://www.top500.org/news/lineshine-debuts-no-1-top500-enters-new-global-exascale-era/
- [2026-06-26] JUPITER's tech page was last updated 26 June 2026 and says installation concludes in 2026. The Jülich FAQ says exascale systems "JUPITER and Alice Recoque" are planned "until the end of 2026". The June entry used 4,801,344 cores (Rpeak 1,226 PF), so the Rmax-to-Rpeak ratio is about 81.5%. This leaves headroom to improve on 1.000 on a re-run, so the main downside risk is not re-running, not a performance drop. https://www.fz-juelich.de/en/jsc/jupiter/tech
- [2026-06 / Nov 2025] The record shows JUPITER, Aurora and Frontier raising or holding their Rmax across lists. Frontier was re-measured up to 1.353, El Capitan to 1.809, and Aurora held at 1.012. The June 2026 list repeated them unchanged. The 4→5 step in June came from a new system, not from improved Linpack runs. https://www.hpcwire.com/2025/11/26/top-500-jupiter-joins-the-exascale-club/
- [2026-06-23] The June entry level was 2.66 PF. The gap between No. 5 (1.000 EF) and No. 6 (0.5715 EF) means only a brand-new system or a large unlisted machine can add to the count. TOP500's own text says the list reflects "only systems submitted". Hyperscaler AI campuses are not designed to run HPL and their owners don't submit, so they are not a plausible source of a sixth system. https://www.datacenterknowledge.com/supercomputers/flops-vs-megawatts-who-s-winning-in-2026-supercomputing-
- [2026-06] Reuters, via Tom's Hardware, reports that Dongarra said Chinese centres had said they were not permitted to submit. A TechWire Asia piece describes the June entry as a return after the US tightened chip controls. LineShine was framed as a deliberate public statement. This points to a politically controlled submission decision in Beijing, so a repeat in November is possible but entirely unsignalled. https://www.tomshardware.com/tech-industry/supercomputers/china-tops-the-top500-with-a-cpu-only-supercomputer-ending-el-capitans-reign
- [2026-10-13] The TOP500 homepage still shows the June list as current. Its heading reads "11/2026 Highlights" but the text is the 67th list. No November preview or date was found. https://top500.org/

BOTTOM LINE: Five existing systems are the base case. The causal drivers of a different count are (1) a new Chinese submission, (2) a surprise US system such as Doudna, and (3) JUPITER or another listed system not re-running or running below its June figure. Doudna's own timeline ("end of the year") makes it unlikely to be benchmarked in time.

NOT FOUND: any announced or leaked November 2026 submission; an official 68th-list release date; confirmation that JUPITER, Aurora or Frontier will re-run; evidence that Doudna, Horizon or Alice Recoque have run full-system Linpack; any news of a further Chinese submission.

report: expert view

I found no prediction-market or forecaster number for this question. I searched for Polymarket, Kalshi, Manifold and Metaculus questions on the November 2026 TOP500 exascale count and found nothing relevant. The evidence below is the official baseline plus analyst commentary.

- [2026-06-23] The official TOP500 baseline is 5 systems at or above 1 EFlop/s HPL, up from 4 in the prior list. They are LineShine at 2.198, El Capitan at 1.809, Frontier at 1.353, Aurora at 1.012 and JUPITER Booster at exactly 1.000. https://www.top500.org/news/lineshine-debuts-no-1-top500-enters-new-global-exascale-era/
- [2026-06-23] JUPITER Booster sits exactly on the ≥1,000 PFlop/s line. The rank-4 Aurora is listed at 1,012.00 PFlop/s. The next system down, Eni HPC7, is at 571.5 PFlop/s, so nothing already listed is close to crossing the line. https://www.top500.org/lists/top500/2026/06/highs/
- [2026-06-23] TOP500 says the June list takes the exascale count "from four to five". It also says exascale systems now sit in Asia, North America and Europe at the same time. https://www.top500.org/lists/top500/2026/06/
- [2026-06-25] Chips and Cheese's George Cozma said he had not expected a new No. 1 until Discovery or Fugaku-Next came online. He asked whether China will also submit Sunway OceanLight and a system he calls CNIS. This is a question, not a forecast. https://chipsandcheese.com/p/top500-at-isc26-we-have-a-new-number
- [2026-06-23] The Next Platform strongly suspects China has several more exascale machines it is not discussing. It names OceanLight at NSC Qingdao (about 1.5 EF peak) and Tianhe-3 at NSC Guangzhou (2.05 EF) as known examples. This is an upside-risk view with no date attached. https://www.nextplatform.com/hpc/2026/06/23/amd-and-nvidia-are-neck-and-neck-in-hpc-supercomputing/5260202
- [2026-06 (approximate; date not shown in results)] Tom's Hardware reports that Jack Dongarra has said for years that Chinese researchers told him they were not permitted to submit. It also quotes Addison Snell of Intersect360, who told Reuters he was surprised that China disclosed the result at all. LineShine's submission is read as a deliberate change of posture. https://www.tomshardware.com/tech-industry/supercomputers/china-tops-the-top500-with-a-cpu-only-supercomputer-ending-el-capitans-reign
- [2026-08 (approximate; about 67 days before today)] Data Center Knowledge lists SC26 in November as the next checkpoint. It cites Alice Recoque's installation as still under way and does not claim it will be on the November list. https://www.datacenterknowledge.com/supercomputers/flops-vs-megawatts-who-s-winning-in-2026-supercomputing-
- [2026 (page dates not shown)] The Alice Recoque schedule points away from November 2026. EuroHPC and CEA say installation starts in 2026. SURF says capacity testing and first users come at the end of 2027. https://www.surf.nl/en/themes/research-infrastructure/alice-recoque
- [2026-10-12] Fontanka reports drone attacks on Yandex data centers on Oct 8–11. They may have damaged the Chervonenkis, Lyapunov and Galushkin systems. These sit at ranks 101, 161 and 134, and are only 12–22 PFlop/s. They are irrelevant to the exascale count. https://www.fontanka.ru/2026/10/12/76696055/
- [2026-06-23] HPE calls the five "verified exascale" systems, including three of its own builds. https://www.hpe.com/us/en/newsroom/blog-post/2026/06/hpe-delivers-six-out-of-ten-of-worlds-most-powerful-supercomputers.html

BOTTOM LINE: The only official baseline is 5, and I found no source projecting a different November number. Plausible changes are an upward move if China or another site submits a new or previously unlisted exascale system, or an upward move if a system near the line is re-run. The Alice Recoque schedule makes it an unlikely November entrant, and nothing found suggests a downward move.

NOT FOUND: Any Polymarket, Kalshi, Manifold or Metaculus market on this question. Any named analyst forecast of the November 2026 exascale count. Any announcement of a new Linpack run or submission for SC26, such as Horizon, Doudna, Discovery or a Chinese system. A published SC26 TOP500 release date (SC26 is "in November" per Data Center Knowledge).

report: favorite case

- [2026-06-23] The official June 2026 table, fetched in full, gives the exact Rmax figures. LineShine is 2,198.40 PF/s, El Capitan 1,809.00, Frontier 1,353.00, Aurora 1,012.00 and JUPITER Booster 1,000.00. No. 6 (HPC7) is 571.5 PF/s, and No. 10 (Alps) is 434.9. The margins over 1,000 PF/s are +1.2% for Aurora and 0.0% for JUPITER. The top five were all "still installed at the same sites". The cut-off is a sharp step, with nothing between 571 and 1,000 PF/s. https://www.top500.org/lists/top500/2026/06/

- [2026-09-23] The Register says JUPITER is still "fifth" on the Top500 as of September. SiPearl only began delivering Rhea1 CPUs to Bull on 22 September. They go into a CPU-only partition of 1,300 nodes that is estimated at about 5 PF. This adds almost nothing to HPL. It also shows JUPITER's build-out is still going. JUPITER therefore has no reason to re-run to protect its place, and an HPL re-run of the Booster could not drop it below 1,000 unless the run were worse than June's. Prior runs (June 2025 at 793 PF, Nov 2025 at 1,000) only moved upward. https://www.theregister.com/systems/2026/09/23/europes-fastest-supercomputer-is-finally-getting-its-domestic-silicon-infusion/5298665

- [2026-10-08] The Register's rundown of new US machines points away from a sixth entry in November. Oak Ridge's Discovery is "expected to come online 2029". Argonne's Solstice (100,000 Blackwell GPUs) is AI-oriented. Nvidia is leaning on FP64 emulation, so these Nvidia systems are not built to post large HPL numbers. No US exascale-class HPL system is flagged for the November 2026 list. https://www.theregister.com/hpc/2026/10/08/nvidia-found-1b-under-the-couch-to-help-secure-american-scientific-computing-dominance/5302110

- [2025-11-06] Alice Recoque's own schedule has delivery starting in Q4 2026, 1 EF Linpack in Q2 2027 and operations in H2 2027. The Register's 23 Sept piece lists it as still waiting on Rhea2 and MI430X parts. A partial-system HPL submission in November 2026 would have to be well above 1 EF, which is not possible at that stage. https://numpex.org/wp-content/uploads/sites/6/2025/12/2025-11-06-02-Alice-Recoque.pdf

- [2026-06-24] Tom's Hardware reports that China stopped submitting its fastest systems around 2021 after entity-list additions. It reads the LineShine submission as "a deliberate change of posture". That posture change is the only identified route to a sixth system. I found no 2026 report that any other Chinese site, such as OceanLight or Tianhe-3, will submit. https://www.tomshardware.com/tech-industry/supercomputers/china-tops-the-top500-with-a-cpu-only-supercomputer-ending-el-capitans-reign

- [2026-06-23] The listing note says El Capitan, Frontier, Aurora and JUPITER all "remain" at the same sites. The June list had no churn at the top, with five exascale systems in a settled composition. The favourite outcome is "5 unchanged". https://www.top500.org/lists/top500/2026/06/

- [2026-08-06] Data Center Knowledge lists the November list at SC26 as "the next opportunity" for a new system. Its other 2026 milestones are Meta's Prometheus (gigawatt-scale AI, not an HPL submission) and Alice Recoque's ongoing installation. It does not name a candidate exascale entrant. https://www.datacenterknowledge.com/supercomputers/flops-vs-megawatts-who-s-winning-in-2026-supercomputing-

- [2026-11-15] SC26 runs Nov 15–20, 2026 in Chicago, with exhibits Nov 17–19. The list is normally unveiled at the conference opening, so a release on or just before the Nov 16 resolution date is plausible. Timing is a small resolution risk, not a count risk. https://agentjesse.ai/lists/sc26-supercomputing-2026-companies

BOTTOM LINE: Five is the clear favourite. It is the June baseline, every member of the top five has stayed at the same site, and no new exascale-class HPL system is dated before 2027. The count could only change through a surprise submission (most plausibly Chinese) or a JUPITER or Aurora re-run that comes in below its June figure.

NOT FOUND: Any announced or leaked November 2026 list contents or submission. An official release date for the 68th list. Any news of a second Chinese exascale submission. Any report that a top-five system will skip or lower its HPL result. I did not check the Aurora, Frontier or El Capitan operator pages for plans to re-run HPL.

report: upset case

The upset case is **6 or more**. A sixth exascale system would need a Chinese submission, a hyperscaler submission, or a very early large build. The best evidence-backed route is a second Chinese submission. The case for **4** is weaker, and I found no source pointing to it. I found no source projecting any count other than the June baseline of 5.

- [2026-06-23] **Policy change in China.** Tom's Hardware reads LineShine as a deliberate change of posture by China. It notes that China stopped submitting its fastest machines around 2021. It also says the community has long believed OceanLight and Tianhe-3 were exascale without ever listing them. A policy shift once made could repeat at SC26. I found no reporting that it will. https://www.tomshardware.com/tech-industry/supercomputers/china-tops-the-top500-with-a-cpu-only-supercomputer-ending-el-capitans-reign
- [2021-11] **Prior Chinese numbers.** Reported HPL results were about 1.05 EF for OceanLight at Qingdao and about 1.3 EF for Tianhe-3 in Tianjin. A third system, NSCC Shenzhen phase 2 at about 2 EF peak, was then reported delayed with no HPL figure. LineShine at Shenzhen may be that third system. If so, the two older machines (OceanLight, Tianhe-3) are the unlisted exascale candidates. Either would clear 1,000 PF if submitted. The figures are unofficial and four years old. https://www.hpcwire.com/2021/11/24/three-chinese-exascale-systems-detailed-at-sc21-two-operational-and-one-delayed/
- [2025-10-06] **JUPITER re-run.** Cloudmagazin, citing Heise, says JUPITER is expected eventually to reach about 1.4 EF FP64 with nearly 24,000 GH200 chips. This does not change the count, since JUPITER already counts at exactly 1.000. It does lower the risk of a dip to 4. I fetched this page, but the 1.4 EF figure is a secondary-source estimate. The same page says the June 2025 run used only 22,000 of the 24,000 GPUs because of cooling limits. https://www.cloudmagazin.com/en/2025/10/06/europes-largest-supercomputer-goes-live/
- [2025-10-06] **Second JUPITER module.** The same page says a further JUPITER module (a Booster expansion, which it also calls a cluster) adds at least one unit in 2026. This is a mechanism for a higher JUPITER score, not for a sixth system. https://www.cloudmagazin.com/en/2025/10/06/europes-largest-supercomputer-goes-live/
- [2026-03-17] **Doudna timing.** Berkeley Lab says the full-scale production Doudna system is to deploy "in late 2026". The early-access system, Cech, is only 5.76 PF FP64. Doudna would need to be built and run HPL before the submission deadline, so I see no realistic route to the November list. A June 2027 entry is far more plausible. https://cs.lbl.gov/news-and-events/news/2026/berkeley-lab-takes-major-step-toward-doudna-with-delivery-of-early-access-system-cech/
- [2025-11-17] **Horizon (TACC).** NSF/TACC specify about 300 PF FP64 and production from spring 2026. That is far below 1 EF, so it does not help the upset case. https://tacc.utexas.edu/news/latest-news/2025/11/17/nsf-lccf-horizon-supercomputer-to-power-breakthroughs-for-the-nations-leading-scientists/
- [2026-08-06] **Hyperscaler campuses.** Data Center Knowledge describes hyperscaler AI campuses (for example Meta's Prometheus) as a separate track measured in gigawatts rather than TOP500 FLOPS. Hyperscaler HPL submissions are rare, and I found none announced. This is an unquantified tail. https://www.datacenterknowledge.com/supercomputers/flops-vs-megawatts-who-s-winning-in-2026-supercomputing-
- [2026-06-23] **Evidence against a Chinese follow-up.** HPCwire notes that China ceased submitting from 2019 under sanctions. Only service-provider Lenovo systems in the mid-400s were submitted. LineShine looks like a single headline submission rather than a general reopening. https://www.hpcwire.com/2026/06/23/surprise-chinese-lineshine-takes-number-1-on-top500/
- [2026-06-23] **Evidence against a fourth-place drop.** The top-5 table on TOP500 shows Aurora at 1.012 and JUPITER at 1.000, with no sign of either being withdrawn. The June list also showed Fugaku in 9th place. https://www.storagenewsletter.com/2026/06/29/isc-2026-the-67th-edition-of-the-top500-list-of-the-worlds-most-powerful-supercomputers-unveiled-in-hamburg/

BOTTOM LINE: The strongest upset option is 6 (or more), and it depends almost entirely on a second Chinese submission such as OceanLight or Tianhe-3. Nothing public signals one, and no US or EU build, including Doudna, is on schedule to reach 1 EF HPL for November.

NOT FOUND: any announcement or leak of a Chinese, hyperscaler or other new ≥1 EF HPL submission for SC26; official confirmation that Aurora, Frontier or JUPITER will re-run; a prediction-market or analyst forecast of the count; any 2026 report of Doudna HPL results.

asknews 6 hot + 10 archive · 18766 chars · 3.8s

AskNews bundle (as the forecaster read it)

16 AskNews articles (newest first; 6 from the last 48 h, 10 from the archive). Summaries are AskNews's own; dates are publication dates.

[2026-10-13] Turkey Offers Cheapest November 2026 All-Inclusive Vacation at 1546 PLN
Nowiny — https://nowiny24.pl/7-najtanszych-kierunkow-na-wakacje-all-inclusive-w-listopadzie-tam-zapomnisz-o-jesieni/ar/c3p2-29434691
- According to a ranking published on October 13, 2026, by the Polish news source Nowiny, Turkey offers the cheapest all-inclusive week-long vacation for November 2026, with a TUI flight to Antalya from Warsaw or Katowice starting at 1546 PLN per person.
- The article identifies the following top seven destinations for affordable sun holidays: 1.
- Turkey (Antalya from 1546 PLN; Alanya from 1562 PLN; Avsallar from 1606 PLN). 2.
- Egypt (Sharm El-Sheikh from 1662 PLN via Kompas Tour; from 1667 PLN via Best Reisen). 3.
- Morocco (Agadir from 1778 PLN via Best Reisen). 4.
- Mallorca, Spain (El Arenal from 1829 PLN via Exim Tours). 5.
- Tunisia (Monastir from 1951 PLN via Sun&Fun). 6.
- Southern Cyprus (Penelope Beach from 1989 PLN via Oasis Tours; Ayia Napa from 2017 PLN via Atur). 7.
- Malta (St.
- Paul's Bay from 2029 PLN via Exim Torus).

[2026-10-12] Drone Attacks on Yandex Data Centers Threaten Russia's Top Supercomputers
Фонтанка.ру — https://www.fontanka.ru/2026/10/12/76696055/
- Between October 8 and October 11, 2026, drone attacks targeted three 'Yandex' data centers in Sasovo (Ryazan region), Kaluga, and Vladimir, Russia, potentially damaging three of the country's most powerful supercomputers: 'Chervonenkis' and 'Lyapunov' in Sasovo, and 'Galushkin' in Vladimir.
- According to 'Fontanka.ru', all three systems are included in the TOP500 international ranking of the world's 500 most powerful computing systems, with 'Chervonenkis' being the most powerful Russian supercomputer on the list.
- As of the article's publication on October 12, 2026, 'Yandex' has not confirmed the extent of the damage or provided recovery timelines. 'Chervonenkis', ranked 101st globally with a performance of 21.53 petaflops, consists of 199 compute nodes and 1,592 NVIDIA A100 graphics accelerators. 'Galushkin' holds the 134th place with 16.02 petaflops (1,088 NVIDIA A100 accelerators), and 'Lyapunov' is ranked 161st with 12.81 petaflops (1,096 NVIDIA A100 accelerators).
- These machines, built between 2020 and 2021 using US-made components from NVIDIA and AMD, are primarily used for training artificial intelligence models.
- Eldar Murtazin, a leading analyst at Mobile Research Group, estimates that replacing a system of this scale would cost approximately 5 billion rubles, but sanctions and supply chain restrictions could increase costs by 30-40% while limiting access to the latest technology.
- The article notes that Russia ranked 21st in June 2026 for the number of systems in the TOP500, with five total systems (three from 'Yandex' and two from 'Sber'), significantly trailing the global leader, China's 'LineShine' at 2.2 exaflops.
- The report also highlights other Russian systems, including 'Sber' supercomputers 'Kristofari' (6.67 petaflops, 256th place) and 'Kristofari Neo' (11.95 petaflops, 167th place), and the St.
- Petersburg Polytechnic University's 'Polytechnic' center, which has a peak performance of over 1.4 petaflops and is used for scientific and industrial research, such as the digital twin project for 25 MW gas turbines.

[2026-10-12] Argonne ATPESC 2026: 75 Participants Train on AI, Vibe Coding, and Exascale Supercomputers
HPCwire — https://www.hpcwire.com/off-the-wire/argonne-extreme-scale-computing-workshop-provides-hands-on-experience-with-supercomputers-and-ai/
- From July 26 to Aug. 7, 2026, the Argonne Training Program on Extreme-Scale Computing (ATPESC) convened nearly 1,000 total participants (75 this year) from 47 institutions at the Q Center in St.
- Charles, Illinois, hosted by the U.S. Department of Energy's (DOE) Argonne National Laboratory and its Argonne Leadership Computing Facility (ALCF).
- The two-week program provided hands-on training in exascale computing, AI, and machine learning, featuring expanded tracks on 'vibe coding' (using natural language prompts to guide AI in writing code) and AI/ML workflows.
- Participants accessed leading supercomputers including Aurora and Polaris at ALCF and Frontier at the Oak Ridge Leadership Computing Facility (OLCF).
- Key figures included Bruno Lopez of Metacomp Technologies, who praised the direct access to experts and leadership-class machines; Deborah Stevens, an Argonne assistant computational scientist, who noted the value for non-computer science researchers; and Sharanjeet Kaur, a postdoc in Argonne's Transportation and Power Systems division, who plans to apply in situ visualization and agentic AI workflows to her work.
- Raymond Loy, director of ATPESC, stated the curriculum was adjusted to align with the DOE's Genesis Mission, which aims to use AI to secure American energy dominance and accelerate scientific discovery.
- The program also fostered a global community, with attendees from the U.S., Italy, Spain, India, Brazil, France, Belgium, Ghana, and China.

[2026-10-12] RRB Announces Recruitment for 3,548 NTPC Graduate Posts with Application Deadline in November 2026
Hindustan Times — https://www.hindustantimes.com/education/employment-news/rrb-ntpc-graduate-recruitment-2026-apply-for-3548-posts-at-rrbapply-gov-in-direct-link-here-101791800192383.html
- The Railway Recruitment Board (RRB) has invited applications for 3,548 Graduate-level posts under the NTPC recruitment drive, with applications accepted online via rrbapply.gov.in.
- The last date to apply is November 6, 2026, and the final date for application fee payment is November 8, 2026.
- A correction window is scheduled to open on November 9, 2026, and close on November 18, 2026.
- Eligible candidates must be between 18 and 33 years of age, with age limits calculated as of January 1, 2027.
- The selection process includes a 1st Stage Computer-Based Test (CBT), a 2nd Stage CBT, and Document Verification/Medical Examination.
- The examination fee is ₹500 for general candidates and ₹250 for candidates belonging to SC, ST, Ex-Servicemen, PwBD, Female, Transgender, Minorities, or Economically Backward Class (EBC).
- Fees can be paid via internet banking, debit/credit cards, or UPI.
- Additionally, the article notes that NISM has launched two certification courses on cyber security and cyber resilience for GIFT IFSC.

[2026-10-12] Néstor Lorenzo Eyes Jorge Carrascal's Return to Colombia National Team for November Friendlies
Bolavip — https://bolavip.com/futbol-colombia/el-gran-regreso-que-tendria-la-seleccion-colombia-para-el-mes-de-noviembre
- According to an article published on October 12, 2026, by Bolavip, Colombia's national team coach Néstor Lorenzo is considering a return for midfielder Jorge Carrascal to the 'Selección Colombia' squad for the upcoming November FIFA dates.
- The report notes that the team struggled in October due to a lack of creativity, previously provided by players like James Rodríguez and Juan Fernando Quintero.
- Carrascal, who was carried to the 2026 World Cup but played zero minutes, is currently performing at a high level for Flamengo.
- Post-World Cup statistics for Carrascal show he has played 11 matches, scoring 4 goals and providing 7 assists, directly participating in 11 of his team's goals.
- Colombia has scheduled two friendly matches in November against Chile, which will conclude their 2026 season.

[2026-10-11] 500 by Midnight Podcast: Vegas Concert Postponements, New Hotel Suites, and 2027 Events
Las Vegas Podcast: Five Hundy by Midnight — http://dts.podtrac.com/redirect.mp3/www.fivehundybymidnight.com/fhbm/podcasts/fhbm1024.mp3
- On October 11, 2026, the Las Vegas podcast '500 by Midnight' (Episode 1024), hosted by Tim and Michelle, covered several entertainment and hospitality updates.
- Regarding cancellations, Lionel Richie postponed his Encore Theater shows scheduled for the following weekend due to ongoing health issues, rescheduling them for early January; Barry Manilow also postponed a Saturday show after a Friday performance where he played only two songs due to bronchitis, though he aims to return next weekend.
- Conversely, the band Chicago announced nine new concerts at the Venetian between February 19 and March 6, 2027, with pre-sales starting Monday via Ticketmaster.
- John Mayer added shows at the Sphere, extending his residency through June with 18 total shows, while the band Tesla is scheduled for five shows at the House of Blues at Mandalay Bay in March 2027, with tickets starting at $77.
- In hospitality news, the Palazzo announced 'Milano Suites,' converting 275 top-floor suites into a hotel-within-a-hotel concept with reservations starting in March; amenities include access to a private pool, a Milano Lounge, and select suites featuring Dyson hair care products and Peloton or Tonal fitness equipment.
- At Excalibur, construction permits indicate three new food court options—Wing Zone, Moe's Southwest Grill, and Haagen-Dazs—are expected within the next six to nine months.
- Additionally, Resorts World's 'Vice Den' (Secret Pussy) has launched a 'Dia de los Muertos' themed pop-up running through November 22, and the 'Barbecue World Cup' is scheduled for April 8-11, 2027, behind the Link and Harrah's, featuring 100 teams competing for a $1 million prize.

[2026-10-09] LineShine Supercomputer Achieves Top Rankings on TOP500 and HPCG Lists
Newswise - Latest News — https://www.newswise.com/articles/no-1-on-top500-and-hpcg-lineshine-supercomputer-introduces-a-new-architecture-for-hpc-ai-converged-scientific-computing/
- On October 09 2026, a new exascale supercomputer named LineShine, deployed at the National Supercomputer Center in Shenzhen, China, has returned to the top of the global TOP500 list.
- In the June 2026 rankings, LineShine achieved 2.1984 EFlops of sustained HPL performance, becoming the first CPU-only system on the TOP500 list to exceed 2 EFlops in sustained double-precision performance.
- LineShine also ranked first on the HPCG list with 22.0049 PFlops.
- The system was designed by Professor Yutong Lu and detailed in a paper published in Frontiers of Computer Science titled 'LineShine: Online Acceleration for HPC-AI Converged Scientific Computing.' The architecture features the LX2 CPU, a chiplet-based many-core processor with 304 cores per CPU, integrating SME matrix and SVE vector engines.
- Each compute node contains two LX2 CPUs, 64 GB of HBM, 512 GB of DDR5 memory, and eight network interfaces.
- At full scale, LineShine consists of about 90 compute cabinets, 22,680 compute nodes, and more than 13 million processor cores.
- The system uses the Kylin operating system, Lclang compiler, and STAR-Lance computation acceleration suite.
- Benchmark results show LineShine reached 2.1984 EFlops of sustained FP64 performance on HPL, with 80.37% computation efficiency and 52.07 GFlops/W system energy efficiency.
- It achieved 22.0049 PFlops on HPCG and 7.92 EFlops on the mixed-precision HPL-MxP benchmark, ranking fourth in the June 2026 HPL-MxP list.
- The paper evaluates LineShine through four production-scale scientific applications: PowerLLEL for engineering simulation, CerebroSim for brain science, GLARE-Exa for drug design, and D2AR for Earth observation.

[2026-10-09] LineShine Supercomputer Returns to Top of TOP500 with New CPU-Only Architecture
newswise — https://www.newswise.com/articles/no-1-on-top500-and-hpcg-lineshine-supercomputer-introduces-a-new-architecture-for-hpc-ai-converged-scientific-computing
- Published on October 09 2026, a paper in Frontiers of Computer Science details the 'LineShine' supercomputer, a Chinese exascale system deployed at the National Supercomputer Center in Shenzhen.
- After nine years, LineShine returned to the top of the global TOP500 list in the June 2026 rankings, achieving 2.1984 EFlops of sustained HPL performance and ranking first on the HPCG list with 22.0049 PFlops.
- This makes it the first CPU-only system to exceed 2 EFlops in sustained double-precision performance.
- The system was designed by Professor Yutong Lu, chief designer of the supercomputer, to handle HPC-AI converged workflows.
- Unlike conventional CPU-plus-accelerator architectures, LineShine uses an 'Online Acceleration' CPU-only architecture with its core processor, the LX2 CPU, which features 304 cores per chiplet and integrates SME matrix and SVE vector engines.
- At full scale, the system consists of approximately 90 compute cabinets, 22,680 compute nodes, and over 13 million processor cores.
- The paper highlights four production-scale applications: engineering simulation (PowerLLEL) handling 16.5 trillion grid points; brain science (CerebroSim) simulating 86 billion neurons and 100 trillion synapses; drug design (GLARE-Exa) traversing 10^13 compounds in 22.6 hours; and Earth observation (D2AR) using a 6.3-billion-parameter generative model.
- The authors note that LineShine is not intended to replace GPU-based systems but offers a complementary path for complex, continuous scientific workflows.

[2026-07-05] China's LineShine Tops 2026 Supercomputer Ranking as Europe Expands AI Infrastructure
Business Insider — https://businessinsider.com.pl/technologie/nowe-technologie/europa-na-mapie-superkomputerow-nie-musimy-sie-wstydzic-przed-usa-i-chinami/r022wgq
- In June 2026, at the ISC High Performance conference in Hamburg, the 67th edition of the TOP500 supercomputer ranking was released, revealing a shifting global landscape.
- The Chinese system LineShine, built at the National Supercomputer Center in Shenzhen, claimed the top spot with a performance of 2.198 exaflops, becoming the first system to exceed two exaflops using only non-standard CPUs without dedicated GPU accelerators.
- LineShine displaced the previous leader, the US supercomputer El Capitan from Lawrence Livermore National Laboratory, which fell to second place, while three other US machines occupied the top four spots.
- Europe, while not holding the top four positions, demonstrated significant strength with four systems in the top ten.
- The JUPITER Booster at the Supercomputing Center in Jülich, Germany, became the first European exascale system, ranking 5th.
- Other European systems in the top ten included Eni's HPC7 (6th), HPC6 (8th), and Switzerland's Alps (10th).
- According to the article, Europe possesses over 150 supercomputers in the TOP500 ranking with a combined power of approximately 5.6 exaflops, supported by the EuroHPC Joint Undertaking established in 2018.
- A major development for the continent was Nvidia's announcement to assist in building and upgrading 35 new AI-optimized supercomputers across 23 European countries, aiming to support over three million researchers.
- In Poland, the Helios system at the Cyfronet AGH Academic Computer Center in Krakow ranked 116th globally with a performance of approximately 19 petaflops, while eight Polish systems in total were listed in the TOP500.
- The article highlights that while the US and China compete for absolute leadership, Europe focuses on a broad, distributed infrastructure and increasing integration with commercial technology providers like Nvidia to support applications in climate modeling, health, and AI.

[2026-07-02] Green500 Rankings Show Performance Per Watt Becomes HPC's New Arms Race
HPCwire — https://www.hpcwire.com/2026/07/02/green500-shows-performance-per-watt-is-becoming-hpcs-new-arms-race/
- At ISC 2026, the Green500 rankings announced alongside the TOP500 list highlight a shift in High-Performance Computing (HPC) from raw speed to performance per watt.
- China's LineShine claimed the title of the world's fastest supercomputer, while Europe's DKRZ facilities successfully defended the top three spots on the Green500 with KAIROS, ROMEO, and the Levante GPU extension.
- This stability is unprecedented; Erich Strohmaier, a senior scientist at Lawrence Berkeley National Laboratory and co-founder of the TOP500 rankings, noted, 'I think it has never happened in the Green500 that the top ten has not changed because the Green500 is about technology, not about size...
- Not just the top three stayed the same.
- Actually, the whole top ten in the Green500 did not change.' The top three systems achieved efficiencies of 73.28, 70.91, and 69.43 Gigaflops/Watt respectively, utilizing NVIDIA GH200 Grace Hopper Superchips and advanced cooling technologies.
- This trend reflects the industry's growing constraint on power and cooling, making energy efficiency a critical factor in procurement and system architecture as AI and exascale systems expand.

[2026-06-29] ISC 2026: China's LineShine Tops TOP500 as Sustainability and Productivity Gaps Dominate Hamburg Conference (analysis)
HPCwire — https://www.hpcwire.com/2026/06/29/isc-2026-china-sustainability-and-the-questions-nobody-can-answer/
- At the ISC 2026 conference in Hamburg, China's LineShine system, a CPU-only machine delivering 2.198 exaflops, claimed the top spot on the TOP500 list, marking the first time a Chinese system has held the number one position since 2017.
- This system features a proprietary stack including the ARMv9-based LX2 processor, LingQi interconnect, and Kylin operating system, explicitly optimized for double-precision floating point rather than the mixed-precision AI workloads that dominate the list.
- Conference discussions, particularly a 'Fishbowl Panel' hosted by Intersect360 Research CEO Addison Snell, focused on the sustainability challenges of deploying gigawatt-scale data centers, with Snell highlighting the lack of answers regarding facility obsolescence in four years.
- Rosa M.
- Badia of the Barcelona Supercomputing Center emphasized energy efficiency as 'more important than ever' amidst rising operational costs and resource concerns.
- On the scientific front, Martin Schulz of the Technical University of Munich argued for a future where classical HPC, AI, and quantum computing coexist, while researchers from Sun Yat-Sen University demonstrated a one-kilometer resolution global Earth system model (AP3ESM) running on Chinese exascale systems.
- The conference also addressed the widening gap between system capabilities and user productivity, with Kyle Chard of the University of Chicago presenting the ACADEMY agentic framework and Taina Coleman of the San Diego Supercomputer Center arguing that machines are optimized by those who do not use them.

[… older articles omitted]

Evidence Synthesis ?s · 0 items

Forecast 46.83s · claude-opus-5-5

Rationale

(a) About five weeks remain; the list should come out at SC26 around Nov 16, and nothing is decided yet. (b) The status quo is 5: the June 2026 list has LineShine, El Capitan, Frontier, Aurora and JUPITER (exactly 1.000 EF) at or above 1 EF. All five remain installed, and carried-over entries keep their Rmax. Nothing else is within about 40% of 1 EF, and the known Western builds (Doudna, Alice Recoque, Discovery) are dated 2027 or later. (c) A surprising but plausible outcome is a sixth system from an unannounced submission, most likely a second Chinese machine such as OceanLight or Tianhe-3, or a hyperscaler or NVIDIA-class AI system running HPL. LineShine showed this can happen with about two months' notice. A drop to 4 would need a withdrawal or a JUPITER re-run below its June figure, which runs against JUPITER's record of only upward re-runs. I therefore put about three-quarters on 5, a moderate tail on 6, and small tails on 4 and 7+.

Evidence brief shown to forecaster

## Tool outputs

### resolution_source (SUCCESS)
RESOLUTION SOURCE (the page the question resolves on — treat as the status-quo anchor):
  Current state: The November 2026 TOP500 list has not been published yet. The lists page only goes up to June 2026 (the 67th list, published June 22, 2026). TOP500's June 2026 news release says five systems are at or above 1 Exaflop/s on HPL: LineShine (2.198 EF), El Capitan (1.809), Frontier (1.353), Aurora (1.012) and JUPITER Booster (1.000). It describes this as an increase from four to five.
  Last updated: 2026-06-23 | already resolved: unclear | page matches criteria: True
  Note: The November 2026 list will probably come out at SC26 in mid-November 2026. The current baseline is 5 systems with Rmax ≥1,000 PFlop/s. JUPITER Booster sits right at the threshold ('exactly 1.000 Exaflop/s'), so check its precise Rmax on the November list, because a re-run below 1,000 PF would drop it out. I did not check the exact June 2026 Rmax value in PFlop/s on the list table itself. New exascale entrants, or re-benchmarked systems such as Alps upgrades or further Chinese or US machines, could raise the count.
  Fetched: fetch: https://www.top500.org/lists/top500/ (HTTP 200)
  Fetched: fetch: https://www.top500.org/news/lineshine-debuts-no-1-top500-enters-new-global-exascale-era/ (HTTP 200)

### wikipedia_lookup (SUCCESS)
Wikipedia (3 lookups):

  TOP500:
  The TOP500 project ranks and details the 500 most powerful non-distributed computer systems in the world. The project was started in 1993 and publishes an updated list of the supercomputers twice a year. The first of these updates always coincides with the International Supercomputing Conference in June, and the second is presented at the ACM/IEEE Supercomputing Conference in November. The project aims to provide a reliable basis for tracking and detecting trends in high-performance computing and bases rankings on HPL benchmarks, a portable implementation of the high-performance LINPACK benchmark written in Fortran for distributed-memory computers.

  Exascale computing:
  Exascale computing refers to computing systems capable of calculating at least 1018 IEEE 754 double precision (64-bit) operations (multiplications and/or additions) per second (exaFLOPS)"; it is a measure of supercomputer performance.

### claude_news (SUCCESS)

Raw forecast prompt

You are an elite superforecaster. Estimate the probability of each option for this Metaculus multiple-choice question.

## Question
How many systems on the November 2026 TOP500 list will be exascale?

## Description / Resolution Criteria
## Description
TOP500: [LineShine Debuts at No. 1 as the TOP500 Enters a New Global Exascale Era](https://www.top500.org/news/lineshine-debuts-no-1-top500-enters-new-global-exascale-era/)

`{"format": "bot_tournament_question", "info": {"hash_id": "c5c08653fe61ec51", "sheet_id": "4"}}`

## Resolution Criteria
This question will resolve as the number of systems on the November 2026 [TOP500 list](https://www.top500.org/lists/top500/) which have an Rmax of ≥1,000 PFlop/s.

## Options
- 4 or less
- 5
- 6
- 7 or more

## Research tool outputs (raw; the resolution-source read is the status-quo anchor)
## Tool outputs

### resolution_source (SUCCESS)
RESOLUTION SOURCE (the page the question resolves on — treat as the status-quo anchor):
Current state: The November 2026 TOP500 list has not been published yet. The lists page only goes up to June 2026 (the 67th list, published June 22, 2026). TOP500's June 2026 news release says five systems are at or above 1 Exaflop/s on HPL: LineShine (2.198 EF), El Capitan (1.809), Frontier (1.353), Aurora (1.012) and JUPITER Booster (1.000). It describes this as an increase from four to five.
Last updated: 2026-06-23 | already resolved: unclear | page matches criteria: True
Note: The November 2026 list will probably come out at SC26 in mid-November 2026. The current baseline is 5 systems with Rmax ≥1,000 PFlop/s. JUPITER Booster sits right at the threshold ('exactly 1.000 Exaflop/s'), so check its precise Rmax on the November list, because a re-run below 1,000 PF would drop it out. I did not check the exact June 2026 Rmax value in PFlop/s on the list table itself. New exascale entrants, or re-benchmarked systems such as Alps upgrades or further Chinese or US machines, could raise the count.
Fetched: fetch: https://www.top500.org/lists/top500/ (HTTP 200)
Fetched: fetch: https://www.top500.org/news/lineshine-debuts-no-1-top500-enters-new-global-exascale-era/ (HTTP 200)

### wikipedia_lookup (SUCCESS)
Wikipedia (3 lookups):

TOP500:
The TOP500 project ranks and details the 500 most powerful non-distributed computer systems in the world. The project was started in 1993 and publishes an updated list of the supercomputers twice a year. The first of these updates always coincides with the International Supercomputing Conference in June, and the second is presented at the ACM/IEEE Supercomputing Conference in November. The project aims to provide a reliable basis for tracking and detecting trends in high-performance computing and bases rankings on HPL benchmarks, a portable implementation of the high-performance LINPACK benchmark written in Fortran for distributed-memory computers.

Exascale computing:
Exascale computing refers to computing systems capable of calculating at least 1018 IEEE 754 double precision (64-bit) operations (multiplications and/or additions) per second (exaFLOPS)"; it is a measure of supercomputer performance.

### claude_news (SUCCESS)

## Question dates and conventions
Dates: opened 2026-10-13; forecasting closes 2026-10-13; resolves 2026-11-16. Only events AFTER the open date can satisfy the criteria; assume the event has NOT happened unless a source shows it has. A question phrased 'before <date>' or 'by <date>' asks whether it happens between now and that date. The deadline for the outcome is the RESOLUTION date, not the close date.

## Recent news (AskNews, raw article summaries, newest first — read these directly)
16 AskNews articles (newest first; 6 from the last 48 h, 10 from the archive). Summaries are AskNews's own; dates are publication dates.