Scan: AI data center in space by December 31, 2027?

1. Planner Opus 5

Tools chosen: polymarket_direct claude_news gdelt_news polymarket_related kalshi_related wikipedia code_execution

Sub-questions (Fermi decomposition)

What is the current Polymarket price and price history for 'AI data center in space by December 31, 2027'?
Which announced orbital data center missions (Starcloud/Starcloud-2, Google Project Suncatcher with Planet Labs, Axiom Space AxDC, SpaceX Starlink compute, Blue Origin) are scheduled to launch before Dec 31, 2027, and what are their stated launch dates?
How many AI accelerators (H100/H200/GB200/TPU-equivalent) will each planned mission actually carry — does any single announced spacecraft meet or exceed the 100-accelerator threshold?
What has already been launched (e.g., Starcloud-1 in Nov 2025 carried a single H100) and does it fall short of the resolution criteria?
What is the historical base rate of schedule slip for first-of-kind commercial satellite/space demonstration missions announced ~2 years out?
Are there cross-venue markets (Kalshi or other Polymarket markets) on space data centers or Starcloud/Suncatcher that provide an independent price signal?

Planner reasoning

This is a Polymarket question about a technology milestone with a specific, fairly demanding threshold (≥100 data-center-grade AI accelerators in orbit by end-2027). The key is mapping known programs (Starcloud/NVIDIA, Google Project Suncatcher, Axiom Space, SpaceX/Starlink compute, Blue Origin) against their announced launch dates and chip counts, then applying aerospace schedule-slip base rates. Market price is the primary anchor.

2. Research

tool	status	time	items	summary
polymarket_direct	OK	0.4s	1	## This Market's Polymarket Data AI data center in space by December 31, 2027? - Current price (probability): 17.50% - 7-day price change: -7.00% - 30-day price change: -7.00% - Total volume: $21,128 (USD notional) - Price range: 17.50% - 27.00% - Data points: 91 days
claude_news	OK	28.7s	19	## Key Findings: AI Data Centers in Space (by Dec 31, 2027) Starcloud (NVIDIA-backed) - Starcloud-1 launched (Nov 2025) and successfully separated from the SpaceX rocket; the ~60kg satellite features the first Nvidia H100 GPU in space. This is a single-GPU test satellite already operational —
gdelt_news	OK	184.4s	40	GDELT: 40 articles across 4 queries (lookback=120d). 'orbital data center launch GPUs': 10 hits \| 'Starcloud satellite H100 GPU': 10 hits \| 'Google Project Suncatcher TPU satellite 2027': 10 hits \| 'space data center Axiom SpaceX Starship compute': 10 hits
polymarket_related	OK	3.8s	0	Scanned 100 active Polymarket markets, kept 0 matches. keyword 'space data center': 0 markets \| keyword 'orbital data center': 0 markets \| keyword 'SpaceX': 0 markets \| keyword 'AI data center': 0 markets
kalshi_related	OK	3.7s	3	3 related markets / summaries. keyword 'space data center': ok \| keyword 'orbital data center': ok \| keyword 'SpaceX launch': ok
wikipedia	OK	0.2s	4	Fetched 4 Wikipedia entries (1 missing pages).
code_execution	OK	90.7s	0	## Monte Carlo Findings — AI Data Center in Space by Dec 31, 2027 Model setup: 4 candidate programs (Starcloud‑2/3, Google Suncatcher, Axiom AxDC, SpaceX), each modeled with (i) a schedule-slip probability (50–80% range, calibrated to first-of-kind aerospace/tech program base rates), (ii) a log

3. Evidence Brief Sonnet, 7295 chars

# Current state No single announced mission yet combines a near-term (pre-2027) launch date with a ≥100-accelerator payload; Starcloud-1 (1x H100, Nov 2025) and Axiom's ISS/Kepler compute nodes (Jan 2026) already flew but fall far short of the 100-accelerator threshold. Resolution requires a *future* launch meeting the bar — Starcloud-2/3, Google Suncatcher, Axiom's optically-interconnected node, or a SpaceX/Blue Origin filing turning into hardware — all still pending as of the research cutoff (mid/late 2026). # Timeline of key events - 2025-11: Starcloud-1 launches with single Nvidia H100 GPU, reaches orbit and operates (confirmed, DCD). - 2025-09: Axiom Space deploys a data-center node to the ISS — demo-scale (confirmed). - 2026-01-11: SpaceX Falcon 9 carries Axiom/Kepler commercial compute nodes to LEO, first "commercial service" compute offering (confirmed). - 2026-03: Starcloud raises $170–250M Series A/B round at up to $2.3B valuation (Y Combinator's fastest unicorn) to fund Starcloud-2/3 (confirmed, TechCrunch/GeekWire). - 2026-03-19: Blue Origin files FCC application for "Project Sunrise," ~51,600 data-center satellites (confirmed filing; no hardware date) (reported). - 2026-06: Google/SpaceX sign $920M/month compute deal; separately Google's Project Suncatcher paper (with Planet Labs) targets launch of 2 prototype TPU satellites "by early 2027" (reported). - 2026-06: Startup "Orbital" raises $5M pre-seed targeting 100,000+ future LEO AI satellites — very early stage (reported). - 2026-08: Starcloud raises additional $250M as "launch options dry up," citing Starship/rideshare bottlenecks (reported, TechCrunch/SiliconAngle). - Planned/unconfirmed: Axiom optically-interconnected ISS node "in 2027"; Starcloud's Vera Rubin space chip flight "late 2028" (post-deadline, reported). # Event Will an orbital "AI data center" (spacecraft carrying ≥100 data-center-grade AI accelerators, for compute/cloud/AI services) be successfully launched into Earth orbit by Dec 31, 2027? # Outcomes to forecast Yes / No # Kalshi market anchor No direct Kalshi YES price was returned by kalshi_direct in this research pull (tool output absent/not provided). Kalshi_related search surfaced only tangential markets (nuclear data center 53%, SpaceX-Mars 21%, Blue Origin-vs-SpaceX Moon 62%) — none is this ticker. **Treat Kalshi price as unknown; Polymarket (17.5% YES, down from 27% high, -7% over 7d and 30d, $21k volume, 91 days of data) is the best available cross-venue anchor.** # Sub-question answers 1. **Polymarket price/history** — Currently 17.5% YES, declined steadily from a high of 27%; -7pp over both 7d and 30d; thin volume (~$21k total). [polymarket_direct] 2. **Announced missions/dates** — Starcloud-2 (multi-GPU incl. Blackwell) and Starclock-3 (200kW, Starship-launched) targeted but Starship commercial access likely not until 2028-29 per CEO; Google Suncatcher/Planet Labs targets 2 TPU satellites "by early 2027"; Axiom's optical ISS node "in 2027"; SpaceX/Blue Origin have only FCC filings (1M and 51,600 satellites respectively), no launch dates. [claude_news] 3. **Accelerator counts** — No confirmed mission currently states a firm ≥100-accelerator manifest with a committed pre-2027 launch date; Starcloud-3 is designed for large scale but its enabling next-gen chip (Vera Rubin) isn't expected to fly until late 2028. [claude_news] 4. **Already launched** — Starcloud-1 (1 H100) and Axiom/Kepler nodes (general compute, not GPU-dense) both fall well short of the 100-accelerator bar. [claude_news, DCD] 5. **Base rate of schedule slip** — Not directly sourced; Monte Carlo model assumes 50-80% slip probability for first-of-kind aerospace/tech programs, consistent with general industry pattern (not empirically cited in raw research). [code_execution] 6. **Cross-venue markets** — Polymarket has no other related markets (0 matches on relevant keywords); Kalshi has no matching market, only adjacent space/data-center markets. [polymarket_related, kalshi_related] # Key facts (high-confidence, factual) 1. [DCD] Starcloud-1 (1x H100) launched Nov 2025, operating in orbit — below threshold. 2. [claude_news] Axiom deployed ISS compute node Sep 2025; Kepler/Axiom commercial nodes launched Jan 2026 — below threshold. 3. [TechCrunch/GeekWire] Starcloud raised $170-250M (2026), valued ~$2.3B, funding Starcloud-2/3. 4. [DCD/Google] Google Suncatcher + Planet Labs targets 2 prototype TPU satellites "by early 2027." 5. [newspaceeconomy] Blue Origin filed FCC application (Mar 2026) for 51,600-satellite "Project Sunrise" — no launch date. 6. [TechCrunch, Aug 2026] Starcloud cites "launch options drying up," a friction/slip risk factor. # Cross-market signals - Kalshi related: no direct match; adjacent space markets show mixed sentiment (Mars landing 21%, Moon race 62%). - Polymarket: 17.5% YES, declining trend, low volume/liquidity — a fairly bearish, thinly-traded signal. - Sportsbook implied: none available. # Analyst opinions and speculation - Starcloud CEO: expects Starcloud-3 to be first "cost-competitive" orbital data center, but Starship commercial access likely not until 2028-29 — undercuts near-term feasibility. [claude_news] - Peter Zeihan (via Yahoo Finance): calls space data center idea "so many levels of dumb," reflecting skeptic camp. [gdelt_news] - Insurance industry (CNBC) treating orbital data centers as a nascent insurable risk class — signals growing but early-stage industry seriousness. [gdelt_news] # Directional lean per outcome - **Yes**: Strong capital inflow to Starcloud ($250M+), explicit 2027 targets from multiple players (Google, Axiom, Starcloud), NVIDIA backing, Monte Carlo model estimates ~72-83% probability given Starcloud's near-term trajectory. - **No**: Polymarket pricing only 17.5% (sharply below model estimate — large disagreement), Starship/launch bottlenecks explicitly cited, historical first-of-kind aerospace slip rates are high, next-gen space chip (Vera Rubin) not expected until late 2028 (post-deadline), and 100-accelerator threshold is a hard, specific bar no announced near-term mission has explicitly confirmed meeting. # Gaps / unknowns - No actual Kalshi YES price captured for this ticker in the research — a material gap given it's supposed to be the primary anchor. - No mission has publicly confirmed a specific accelerator count ≥100 with a committed pre-2027 launch date; feasibility of Starcloud-2/3 hitting both schedule and scale simultaneously is unverified. - Large disconnect between Polymarket (17.5%) and independent Monte Carlo model (~72-83%) is unresolved — could reflect market skepticism the model underweights (e.g., "consensus of credible reporting" resolution ambiguity, stricter interpretation of "data-center-grade" or "orbit" is more delayed than models assume). # Calibration anchors - Polymarket current price: 17.5% YES (declining trend) — key external anchor given missing Kalshi price. - Monte Carlo/base-rate model: ~72-83% central estimate, 56-93% range — reflects optimistic capacity/announcement momentum but likely overweights announced dates vs. real slip risk. - Precedent: first-of-kind satellite/aerospace demo programs announced ~2 years out historically slip 50-80% of the time (per model assumptions, not directly sourced).

4. Forecasters Opus 5 + GPT-5.5, in parallel

claude-opus-5

60% conf

Yes	13%
No	87%

No announced mission publicly commits to a ≥100-accelerator payload with a pre-2027 launch: Starcloud-1 flew a single H100, Axiom/Kepler nodes are general-purpose demo scale, Google's Suncatcher targets only 2 prototype TPU satellites by early 2027, and Starcloud-3's enabling Vera Rubin chip isn't expected to fly until late 2028. Starcloud itself flags launch options 'drying up' with Starship commercial access likely 2028-29, a hard physical bottleneck for a 200kW-class payload. Polymarket's 17.5% YES, trending down from 27% on thin volume, is the best cross-venue anchor absent a Kalshi price; the Monte Carlo ~75% estimate appears to overweight announced target dates versus first-of-kind aerospace slip rates. I shade slightly below Polymarket because the specific 100-accelerator bar is stricter than the loose 'space data center' narrative that thin market may be pricing, but keep meaningful weight on a surprise dense Blackwell-class Starcloud-2 manifest in 2027.

gpt-5.5

58% conf

Yes	16%
No	84%

With no Kalshi-direct price available, I anchor on the best cross-venue signal: Polymarket at 17.5% YES, declining from 27% on thin volume. The announced-mission evidence points toward No because Starcloud-1 and Axiom/Kepler are already launched but far below the ≥100-accelerator threshold, while no confirmed pre-deadline mission has publicly committed to ≥100 data-center-grade accelerators. Schedule evidence also points toward No because Starcloud’s larger plans appear dependent on constrained launch access and Starship availability likely slipping to 2028-29, while Vera Rubin space-chip timing is cited as late 2028. The main Yes case is real capital inflow and multiple prototype targets in 2027, but these look more like demo-scale deployments than a documented qualifying orbital AI data center by Dec. 31, 2027.

5. Devil's Advocate Sonnet

Both forecasters anchor almost entirely on Polymarket's thin, low-volume (~$21k) price without adequately discounting for its poor liquidity and possible mispricing given the brief explicitly flags "sharply below model estimate — large disagreement" as unresolved; neither seriously reconciles why the Monte Carlo model diverges so much (72-83% vs 17.5%) beyond a cursory dismissal, leaving the anchor choice under-justified.

Neither forecast meaningfully engages with the possibility that "resolution ambiguity" itself could favor Yes — e.g., a looser consensus interpretation of "data-center-grade" or aggregation of a satellite constellation reaching 100 accelerators cumulatively (relevant given Blue Origin's 51,600-satellite filing and Orbital's 100,000+ satellite ambitions), which the brief flags as a real unknown affecting resolution criteria.

Both treat the 2 TPU Suncatcher satellites and Axiom's optical node as clearly sub-threshold, but neither explicitly quantifies the tail scenario where Google or another well-capitalized player (Google/SpaceX $920M/month compute deal) accelerates a denser payload — the brief's "Yes" case cites "explicit 2027 targets from multiple players" that both forecasts underweight in favor of the Starcloud-centric slip narrative.

Both forecasts converge very tightly (0.130 vs 0.160) despite claiming independent reasoning, which risks anchoring bias/herding on the same Polymarket figure rather than independently stress-testing the Monte Carlo model's 56-93% range — the wide range itself suggests more uncertainty than the tight ~13-16% both land on.

Neither forecast explicitly addresses the base-rate evidence quality caveat: the "50-80% slip probability for first-of-kind aerospace/tech programs" is explicitly labeled "not empirically cited in raw research" (i.e., assumption-driven, not sourced), yet both treat this slip-risk framing as solid evidentiary support for No.

Both dismiss the Starcloud CEO's 2028-29 Starship timeline as decisive, but this pertains specifically to Starcloud-3's 200kW payload; smaller rideshare-launched deployments (e.g., a scaled Starcloud-2 with Blackwell GPUs, not dependent on Starship) could plausibly hit 100 accelerators earlier via non-Starship launch vehicles — this alternative pathway is underexplored in both rationales.

6. Aggregator Opus 5

16% (59% confidence)

Yes: 16% No: 84%

The best available anchor (Polymarket 17.5% YES, declining from 27%) aligns with the fundamentals: no announced mission publicly commits to a ≥100 data-center-grade-accelerator payload with a pre-2028 launch, while everything actually flown (Starcloud-1's single H100, Axiom/Kepler nodes) is orders of magnitude below the bar. Starcloud's own launch-access bottleneck (Starship commercial availability 2028-29) and Vera Rubin space silicon in late 2028 both push the dense-payload pathway past the deadline; Google Suncatcher's two prototype TPU satellites and Axiom's optical node are explicitly demo-scale. The critique is fair that the ~75% Monte Carlo estimate is under-reconciled and that a rideshare-launched, Blackwell-dense Starcloud-2 or a loose resolution reading (constellation aggregation, generous 'data-center-grade' interpretation) offers a real tail — this justifies not shading below the market and keeping YES in the mid-teens rather than sub-10%. Given the assumption-driven (not sourced) slip base rate and thin market liquidity, I widen slightly toward the upper end of the two forecasts rather than the lower.

Pipeline Timing

Total pipeline time: 274.4s

Per-tool research timings shown in the Research section above.