← back to rankingVG · Venture Global, Inc.
Oil & Gas Midstream · mkt cap $31.8B · calls: Q1 FY2026 vs Q4 FY2025
46.0 conviction · conf-adj 46
conf 6/10 partial
enthusiasm:21.0 · trend:8 · quantifies:5 · impact:0 · under_radar:5 · credibility:0 · business_impact:8 · disruption:0 · commitment:-4 · confirmation:3
Enthusiasm latest 7 / prev 6 (rising)
VG's AI thesis is industrial rather than customer-facing: the company deploys data science and AI tools on a sensor network now exceeding 800K–1M readings every 10 seconds to push LNG throughput 40%+ above nameplate capacity, which management credits as the single largest driver of its margin advantage since that incremental output is essentially 'free' against already-funded fixed costs. The claims are operationally grounded—tied to specific FERC capacity filings, declining $/MMBtu metrics, and a growing cargo record—lending them above-average credibility relative to typical AI enthusiasm, though no standalone AI revenue or cost-savings line item is disclosed. Enthusiasm is rising modestly between the two calls, most notably in Sabel's Q1 2026 assertion that 'data acquisition is likely as valuable or more valuable than the rest of the business.'
GROUNDED NEXT-FY IMPACT vs CONSENSUS
Grounded on actual base — revenue $13.8B · net income $2.7B · net margin 19.6% · diluted EPS 0.86
These are next-fiscal-year annual uplift estimates, not next-quarter numbers.
Aggregate next-FY est. rev uplift: 1.54% · next-FY EPS uplift: 2.8% · vs analysts: inline · priced in: medium (model's call-read: high; verdict above is the hard-data one used for ranking) · confidence: 6/10
| Claim | Figure | Arithmetic | Next-FY Rev % | Next-FY EPS % |
|---|
40%+ production above nameplate, attributed to AI/data science team (CP + Plaquemines)
revenue | 40%+ above nameplate | Nameplate 10+20=30 MTPA; actual 42 MTPA; AI-attributed throughput ~28.6% of output (40/140) = ~$3.93B of $13.77B rev — but this is a PRODUCTIVITY/throughput effect already realized and embedded in the FY25 base and FY26 ramp consensus, not new list-price revenue. For NEXT-FY INCREMENTAL portion: extra volume is largely under existing nameplate contracts, so monetization is contract-limited (~25%); ~30% of that still incremental vs FY25 embed -> ~$212M rev = 1.54%. Incr NI @25% incr margin (mgmt 'higher margins on extra')=$53M; eps=100×53/2,697=1.97%. (X sized the FULL in-base contribution at 28.6% rev/58.3% EPS; that is total, not next-FY incremental.) | 1.54 | 1.97 |
O&M ~30% below industry average, partly from data capture/analysis
cost · soft | ~30% below industry average | Mgmt attributes the 30% to a COMBINATION (modular design + data capture + continuous learning); AI's isolable share is not separable. Even deriving an implied industry cost from the disclosed $0.30/MMBtu (0.30/0.70=$0.4286), attributing a specific fraction (e.g. 33%) to AI is invented, not disclosed. Combination % with no AI-isolable base -> soft. | | |
500k -> 800k-1M data collection points per 10s
other · soft | 500k -> ~1M data points / 10s | Operational input-scale metric; enabler of the other claims, no $, volume, or margin base to convert to rev/EPS. Not sizeable. | | |
Plaquemines cost/MMBtu $0.44 (nameplate) -> <$0.30 (full capacity)
cost | $0.44 -> <$0.30 /MMBtu (~$0.14 unit-cost cut) | DISCLOSED Δ≥$0.14/MMBtu. Plaquemines 20 MTPA at 140%=28 MTPA=1,456M MMBtu/yr; pretax saving=1,456×0.14=$204M; after-tax×0.79=$161M steady (5.97% EPS). Next-FY: ~40% incremental vs FY25 capture=$64M after-tax -> eps=2.39%. NOTE: this is the cost-side mechanism of claim 1's above-nameplate margin — partially overlaps; an overlap haircut is applied at the aggregate to avoid double-count (X excluded it entirely as full double-count). | 0 | 2.39 |
Assumptions: Next FY = FY2026 (ending 2026-12-31). Nameplate Calcasieu Pass 10 + Plaquemines 20 = 30 MTPA; 40% above -> 42 MTPA; 52 MMBtu/tonne. The 40%+ throughput (~28.6% of rev, ~$3.93B) is AI's TOTAL operational contribution but is already in the FY25 base AND the FY26 Plaquemines-ramp consensus, so only the next-FY INCREMENTAL slice is sized. Contract-limited monetization 25%; 30% of physical incremental still incremental vs FY25; incremental net margin 25% (mgmt 'higher margins', vs 19.6% blended). Tax 21%. O&M 30%-below-industry is a combination metric with no AI-isolable base -> soft. Plaquemines $/MMBtu saving overlaps claim 1's margin mechanism; ~35% overlap haircut applied between claim 1 (throughput margin) and claim 4 ($/MMBtu) at aggregate, and claim 2 excluded -> est_eps_uplift_pct ~2.8%. EPS denominator = reported FY25 NI $2,697M (company base as provided; FY25 consensus NI $2,352M is lower).
Top line: AI/data science drives the 40%+ above-nameplate production (~28.6% of current revenue, ~$3.93B of $13.77B) — a large, genuinely operational adopter contribution, but it is already realized and embedded in the FY25 run-rate AND in consensus FY26 revenue ($17.9B, +30%, the Plaquemines ramp). The NEXT-FY incremental, contract-limited monetization is narrow: ~$212M (1.54%). No incremental above-consensus AI revenue is quantified.
Bottom line: Hard cited math: Plaquemines $0.14/MMBtu spread on ~1.46B MMBtu = ~$204M pretax / $161M after-tax steady (next-FY incremental ~$64M, 2.39% EPS); above-nameplate throughput at higher margin adds ~$53M NI (1.97% EPS). With claim 2 (combination O&M) soft and a ~35% overlap haircut between the throughput-margin and $/MMBtu mechanisms, next-FY adopter EPS uplift ~2.8% (~$75M NI) on $2.70B FY25 NI — modest vs consensus FY26 NI step-up of ~$1.2B. The full ~58% in-base EPS contribution X cites is structural and already priced, not a forward surprise.
Consensus FY26 rev $17.9B (+30%) and EPS ~$1.45 (+~70%) are driven by the Plaquemines ramp, which inherently embeds above-nameplate (AI-attributed) throughput — analysts model actual run-rate output, not nameplate. So the AI contribution sits INSIDE consensus, not on top. FY25 actual NI ($2.70B)/EPS ($0.86) already exceed FY25 consensus, suggesting throughput optimization is partly in the base. Next-FY incremental AI math is small (~1.5% rev, ~2.8% EPS); the only un-priced upside (CP2 engineered throughput, 6.4 MTPA bolt-ons, fleet leverage 'in nascency') is explicitly unquantified -> soft. Hence high priced-in, inline.
MODEL CONSENSUS (impact)
partial
Both verdicts agree (inline, high priced-in, conf 6). Disagreement was framing: X sized total in-base AI contribution, Y the next-FY increment; took Y's methodologically-correct next-FY numbers.
Conflicts reconciled
- est_rev_uplift_pct: X=28.6 vs Y=1.54 -> used 1.54 because the task requires the NEXT-FY incremental portion; X's 28.6% is the total in-base contribution already embedded in consensus
- est_eps_uplift_pct: X=58.3 vs Y=3.7 -> used 2.8 (next-FY incremental, claims 1+4 with ~35% overlap haircut, claim 2 dropped as soft)
- claim2 soft: X=true vs Y=false -> used true (conservative): isolating AI's share of the combination O&M metric requires invented attribution
- claim4 aggregation: X=null/double-count vs Y=2.39 -> kept 2.39 but applied overlap haircut at aggregate to address X's double-count concern
| Field | Opus 4.8 | GPT-5.5 |
|---|
| Rev uplift % | 28.6 | – |
| EPS uplift % | 58.3 | – |
| Priced in | high | – |
| vs analysts | inline | – |
| Confidence | 6 | – |
| Top line | AI/data science is credited with the 40%+ above-nameplate production = ~28.6% of current revenue (~$3.93B of $13.77B) — a large, genuinely operational adopter contribution. But it is already realized and recurring, so consensus FY2026 revenue of $17.9B (+30%, the Plaquemines ramp) already embeds above-nameplate output. No incremental, above-consensus AI revenue figure is quantified for the next fiscal year. | – |
| Bottom line | Because the extra volume is funded by nameplate contracts, it flows at high incremental margin (~40% net assumed; corroborated by Plaquemines unit cost falling $0.44-><$0.30/MMBtu at full capacity). That implies ~$1.57B of NI, or ~58% of current $2.697B NI, is attributable to data-optimized throughput. The ~30% O&M advantage reinforces this but is a combination metric with no isolable AI dollar base (soft). All of this is structural and already in the base — not a forward EPS surprise. | – |
| Reasoning | Consensus FY2026 rev $17.9B (+30% vs $13.77B) and EPS $1.45 (+69% vs $0.86) are driven by the Plaquemines ramp, which inherently embeds the above-nameplate (AI-attributed) production — analysts model VG's actual run-rate output, not just nameplate. So the 28.6% rev / 58.3% EPS AI contribution sits INSIDE consensus, not on top of it. The only un-priced upside (CP2 engineered throughput gains 'hopeful we do better,' 6.4 MTPA bolt-ons, fleet-wide operating leverage 'in its nascency') is explicitly unquantified -> soft. Hence high priced-in, inline vs expectations. | – |
Rows highlighted where the two models disagreed.
QUANTIFICATIONS
Production above nameplate capacity attributed to data science/AI: 40%+ above nameplate (Current (Q1 FY2026), realized across Calcasieu Pass and Plaquemines, both)
“the value of that extra 40-plus percent production capacity because it's largely funded by the contracts that we signed for the nameplate capacity has higher margins and higher value for us. So it's all driven by our data science team”
O&M cost advantage vs. industry average, attributed partly to 'massive data capture and analysis': ~30% below industry average (Current (Q4 FY2025), bottomline)
“the combination of structural benefits from our modular approach, massive data capture and analysis, and our unrelenting focus on continuous learning and improvement translates into superior LNG production and project-level operating and maintenance costs that are currently about 30% below industry averages”
Data collection points per facility (operational AI input scale): 500,000+ per 10-second interval (Q4 2025) → 800,000–1,000,000 per 10-second interval (Q1 2026) (Q4 FY2025 → Q1 FY2026 (growing), bottomline)
“I think we are now capturing over 500,000 data collection points every 10 seconds between Calcasieu Pass and Plaquemines [Q4 2025]... we're somewhere between 800,000 plus and 1 million data collection points between our 2 facilities [Q1 2026]”
Cost per MMBtu at Plaquemines nameplate vs. full capacity (benefit flows through data-optimized throughput): ~$0.44/MMBtu at nameplate → below $0.30/MMBtu at full capacity (Current trajectory (Q1 FY2026), bottomline)
“at Plaquemines, where on a nameplate basis, we're at roughly $0.44 per MMBtu and at full capacity, we're in below $0.30 per MMBtu area”
PAST (realized)
- Q4 FY2025 — Sabel: 'We leveraged data generation and analysis lessons learned from Calcasieu Pass to further increase production capacity at Plaquemines.'
- Q4 FY2025 — Sabel: 'We have achieved extraordinary dividends from it [data science/AI].'
- Q1 FY2026 — Sabel: data-driven AI work 'has resulted in the output that you're seeing today' (record 130 cargoes in Q1, production above nameplate).
CURRENT (now)
- Q1 FY2026 — Sabel: 'We use AI tools, both to optimize production' and to identify 'the most valuable data to stream' to manage storage costs.
- Q1 FY2026 — Sabel: 'we're somewhere between 800,000 plus and 1 million data collection points between our 2 facilities' streamed every 10 seconds.
- Q1 FY2026 — Sabel: data/AI work 'allows us to run very sophisticated models and simulations and experiments on operational improvements and optimizations.'
- Q4 FY2025 — Sabel: 'we have a large data science team and AI programmers that consume that data and incorporate it into our operations and our process design.'
- Q4 FY2025 — Sabel: capturing 'over 500,000 data collection points every 10 seconds between Calcasieu Pass and Plaquemines.'
- Q4 FY2025 — Sabel: 'We view our LNG infrastructure as technology assets that we are constantly optimizing to be safer, faster, and more efficient.'
FORWARD (guidance)
- Q1 FY2026 — Sabel: 'the value is increasing from that data' as cargo count surpasses 900 and data volume grows.
- Q1 FY2026 — Sabel: data science learnings 'was integrated into our control rooms and to our design changes and execution not [at] CP2 and the bolt-ons,' meaning AI-optimized designs are embedded in the next ~20+ MTPA under construction.
- Q1 FY2026 — Sabel: 'I'm hopeful we do better than that at CP2 because we engineered in some fairly significant improvements in throughput capacities into CP2 compared to Plaquemines and from lessons learned and from our data.'
- Q4 FY2025 — Sabel: data-driven approach will support bolt-on expansions (6.4 MTPA each at CP2 and Plaquemines), further driving down per-unit O&M costs via fleet-wide operating leverage 'in its nascency.'
TRACK RECORD — PROMISE vs DELIVERY
—/100 (no quantified promises) no-quantified-promises 6 calls reviewed
Across four Venture Global calls (Q1–Q4 FY2025 and Q1 FY2026), management cites data capture, analysis, and operational learning for efficiency but never states a dated numeric AI/ML/automation target (e.g., % cost cut, autonomy milestone, model-driven productivity by year). LNG/cargo/EBITDA/COD guidance dominates and is outside the AI promise scope for this audit.
PRICED-IN (REFINED)
MEDIUMEst. revisions rising · Fwd P/E 15.8 · EV/Sales 4.4x
AI claim maps to Liquefied Natural Gas, Product and Service, Other
Analyst sentiment is migrating up (buy/strong-buy counts rose from 8–9 to 10 since Feb while holds stayed at 9) and price targets stair-step higher (lastYearAvg 13.83 → lastQuarterAvg 15.63 → lastMonthAvg 16.4), with FY2026 consensus embedding a sharp EPS step-up to 1.45 from 0.83. That rising revision momentum suggests much of the operational/ramp upside is already in the numbers, though valuation is only moderately stretched (~15.8x forward P/E, ~4.4x EV/Sales vs ~12x TTM P/E), not a clear premium bubble. Any AI-driven efficiency or throughput gains would most plausibly flow through the LNG line (and secondarily Other services), so the mixed picture—rising estimates but not rich multiples—supports a medium priced-in verdict rather than low or high.
COVERAGE — ENTHUSIASM TRAJECTORY + CATALYSTS
4Q2 FY20217Q3 FY20211Q1 FY20251Q3 FY20254Q4 FY20251Q1 FY2026
AI enthusiasm across 6 calls — trend ↘ falling
Early calls cite product AI wins; VG calls are silent except brief operational data-analytics in Q4 FY2025.
RECENT AI CATALYSTS & NEWS
BUSINESS IMPACT - QUALITATIVE MATERIALITY
7/10 qualitative impact material near-term · mixed evidence
Where AI matters: LNG throughput & O&M unit economics
Industrial AI on 800k–1M sensor streams every 10s is credibly tied to 40%+ above-nameplate LNG output and sub-industry $/MMBtu O&M, lifting margin on largely fixed-cost assets; consolidated next-FY uplift is only ~1.5% revenue and ~3.7% EPS with no AI revenue line and overlap with modular design and ramp already in consensus.
Caveats: 40%+ throughput and O&M edge partly reflect modular design and contract structure, not AI alone; FY25–26 operational outperformance and rising estimates likely embed much of the benefit; Competitors can adopt similar operational analytics without changing LNG demand fundamentals
AI DISRUPTION / CANNIBALIZATION RISK tailwind · 1/10
VG monetizes physical LNG liquefaction and contracted cargoes, not billable knowledge labor; AI optimizes plant throughput and cost rather than commoditizing the commodity or automating away regulated export infrastructure and long-term offtake economics.
OPTIONS / MARKET STRUCTURE
option liquidity: good
proxy inputs — dollar-ADV $347M · beta 0.262736 · px $13.02
source: proxy (no options chain on FMP)
FMP /stable/ exposes no options-chain endpoint on this key, so ATM IV, bid-ask spread and open interest are unavailable. Liquidity below is a PROXY from dollar-ADV, beta and price level (a stand-in for option depth), not measured option-market data.
CONFIRMATION — INSIDERS · 13F · LANGUAGE
Mixed — insiders selling, institutions adding, management language 1/10 hedged.
INSIDERS selling 32 open-market sell(s) vs 0 buy(s) — net distribution
INSTITUTIONS (13F) adding as of 2026-03-31: 154 new / 45 closed positions; 253 increased / 57 reduced; institutional ownership -0.71pp; +135 net 13F holders
MGMT LANGUAGE 1/10 hedged Provided Q1 FY2026 transcript has zero AI/ML/automation references for Venture Global's business; no commit or hedge language to score.
VERBATIM AI QUOTES
“We use AI tools, both to optimize production. But we actually have -- we were able to acquire so much data we use tools to identify the most valuable data to stream and so we can manage our data storage cost.”
— Michael Sabel (CEO), Q1 FY2026
“we're somewhere between 800,000 plus and 1 million data collection points between our 2 facilities.”
— Michael Sabel (CEO), Q1 FY2026
“I think that our data acquisition is likely as valuable or more valuable than the rest of the business.”
— Michael Sabel (CEO), Q1 FY2026
“it's what's enabled us to grow our production capacity from Calcasieu pass where we're for the moment a little bit above nameplate capacity to where we are with Plaquemines and where we're excited to be for CP2, the value of that extra 40-plus percent production capacity because it's largely funded by the contracts that we signed for the nameplate capacity has higher margins and higher value for us. So it's all driven by our data science team, process engineers on that team and was integrated into our control rooms and to our design changes and execution not CP2 and the bolt-ons. It's incredible, incredible value for us.”
— Michael Sabel (CEO), Q1 FY2026
“the magnitude of the data that we stream every 10 seconds on our facilities is immense and it allows us to run very sophisticated models and simulations and experiments on operational improvements and optimizations that have resulted in the output that you're seeing today.”
— Michael Sabel (CEO), Q1 FY2026
“I would say it is almost all driven by the mass amount of data collection that we process. I think we are now capturing over 500,000 data collection points every 10 seconds between Calcasieu Pass and Plaquemines, and we have a large data science team and AI programmers that consume that data and incorporate it into our operations and our process design. We have achieved extraordinary dividends from it, so we are extremely pleased.”
— Michael Sabel (CEO), Q4 FY2025
“the combination of structural benefits from our modular approach, massive data capture and analysis, and our unrelenting focus on continuous learning and improvement translates into superior LNG production and project-level operating and maintenance costs that are currently about 30% below industry averages, and we see room for further improvements.”
— Michael Sabel (CEO), Q4 FY2025
“We leveraged data generation and analysis lessons learned from Calcasieu Pass to further increase production capacity at Plaquemines.”
— Michael Sabel (CEO), Q4 FY2025
“We view our LNG infrastructure as technology assets that we are constantly optimizing to be safer, faster, and more efficient.”
— Michael Sabel (CEO), Q4 FY2025
ANALYST QUESTIONS ON AI
Q (Q1 FY2026, Christopher Robertson (Deutsche Bank)): Mike, in the past, you guys have talked about lessons learned in terms of the construction and commissioning process and about applying those lessons to the future facilities. But I just want to focus more on the operational side here because it's pretty clear that things are going extremely well from an operational perspective at Plaquemine. Can you talk about any operational lessons learned here? Anything gleaned from the data collection efforts that you then apply at CP2 or of future facilities?
A: Sabel deferred first to Thayer on cost structure (high fixed / low variable model, cost per MMBtu declining from ~$0.45–$0.50 toward sub-$0.40 at Calcasieu Pass and below $0.30 at full Plaquemines capacity), then added: 'On the data side of your question, we are super -- remain super excited about it. We...are somewhere between 800,000 plus and 1 million data collection points between our 2 facilities. We use AI tools, both to optimize production...our data acquisition is likely as valuable or more valuable than the rest of the business...it's all driven by our data science team, process engineers on that team and was integrated into our control rooms and to our design changes and execution not CP2 and the bolt-ons...the magnitude of the data that we stream every 10 seconds on our facilities is immense and it allows us to run very sophisticated models and simulations and experiments on operational improvements and optimizations that have resulted in the output that you're seeing today.'
Q (Q4 FY2025, Manav Gupta (UBS)): My first is we recently saw a filing by you in which you basically indicate that you would be in a position to run Plaquemines at 35 MTPA and maybe CP2 at 35. I am trying to understand how you are able to find these incremental volumes in your system. Is it the engineering? Is it the design? Is it operational efficiency? Are the massive data operations that you have set up? Help us understand how these incremental volumes are available in the system.
A: Sabel: 'I would say it is almost all driven by the mass amount of data collection that we process. I think we are now capturing over 500,000 data collection points every 10 seconds between Calcasieu Pass and Plaquemines, and we have a large data science team and AI programmers that consume that data and incorporate it into our operations and our process design. We have achieved extraordinary dividends from it, so we are extremely pleased. We are able to, particularly because we have so many trains that we operate now, experiment with changes in configuration that allow us to fine-tune production.'