Goldman's 67GW Behind-the-Meter Forecast Splits GEV and BE Options Exposure

The ticker matters before the options chain does#

Goldman Sachs has raised the scale of the data-center power problem again, but the most useful options question is not whether electricity demand is growing. It is which listed power market is actually exposed to the part of the forecast that changed. A September 23 summary of Goldman Sachs Equity Research puts global data-center capacity at 217 GW in 2030, up from a prior 168 GW forecast. It also raises expected 2025-30 global data-center power-demand growth to 170% from 117% and lifts the forecast for global behind-the-meter generation to 67 GW from 40 GW. Goldman's own September 1 public discussion independently confirms the 170% power-demand forecast and says its 2030 capacity buildout assumptions had risen by about 25%. It also describes behind-the-meter generation as an increasingly important bridge around grid interconnection delays that can stretch to seven years. That revision does not map evenly into one "AI power" ticker. GE Vernova and Bloom Energy sit on different parts of the same constraint. GEV spans gas turbines, grid equipment and electrification. BE is more concentrated on onsite fuel-cell generation. The same 67 GW behind-the-meter forecast therefore reaches the two companies through different economic mechanisms, and their options markets reflect very different uncertainty and liquidity profiles.

The first correction is mechanical but important.

GE is GE Aerospace. The power-infrastructure company created from the former General Electric portfolio is GE Vernova, which trades as GEV.

That distinction is not cosmetic. GE Aerospace's primary economic exposure is commercial and defense aviation. GE Vernova's Power and Electrification businesses are the units tied to gas generation, grid equipment and the data-center buildout.

For an options article about AI power infrastructure, substituting GE for GEV would break the transmission map before any volatility analysis begins.

Goldman is revising both the amount of power and how it may be supplied#

The 217 GW capacity figure is only the first layer of the forecast.

Goldman's September public discussion says global data-center power demand could rise 170% by 2030 from 2025 levels, compared with a 117% forecast earlier in the year. The firm also says roughly 60% of data-center demand may ultimately be met with natural gas and about 40% with renewables, while grid delays are forcing developers to consider power sources outside the traditional interconnection path.

Behind-the-meter generation is the key structural change.

Instead of waiting for a data center to connect through the utility system, the developer can place generation at or near the site and serve load directly. Goldman describes that approach as a bridge rather than necessarily the permanent architecture, because the grid remains cheaper and more reliable at very large scale once interconnection becomes available.

The September 23 research summary raises Goldman's global behind-the-meter forecast from 40 GW to 67 GW by 2030. That 27 GW revision is more informative for company mapping than the total 217 GW capacity figure because it changes the expected mix of infrastructure required to power the buildout.

GEV reaches the forecast through several layers#

GE Vernova is exposed to the data-center power problem through more than one product category.

Its second-quarter update said Gas Power equipment backlog and slot-reservation agreements had reached 116 GW, with management expecting at least 125 GW by the end of 2026. The company plans to raise annual gas-turbine output from 20 GW in the third quarter of 2026 to 24 GW in 2028 and is implementing actions to reach 30 GW in 2030.

The same update said Electrification had generated more than $5 billion of data-center orders year to date, already more than double the company's total for all of 2025.

That means GEV can participate in both sides of the architecture.

If data centers remain grid connected, transformers, switchgear and other electrification equipment matter. If developers move temporarily or permanently behind the meter, gas turbines can matter. A later grid connection can still create transmission and electrification work even if an onsite plant is used first.

The broad exposure reduces the risk of treating one power architecture as the only path, but it also makes GEV less pure as a proxy for Goldman's specific 67 GW behind-the-meter revision.

BE is narrower, but the link is more direct#

Bloom Energy's economic mechanism is easier to isolate.

Its solid-oxide fuel-cell systems generate power onsite, which places the product directly inside the behind-the-meter category. Goldman has previously identified fuel cells as one of the principal technologies available for data-center onsite generation and estimated that fuel cells could ultimately supply a meaningful portion of incremental behind-the-meter demand.

Bloom also has operating evidence connecting that technology to hyperscale AI infrastructure.

In April, Bloom expanded its agreement with Oracle to support up to 2.8 GW of fuel-cell capacity. The company said 1.2 GW had already been contracted and deployment was underway. In July, Bloom reported quarterly revenue of $1.065 billion, up 166% from the prior-year period, and raised full-year 2026 revenue guidance to $3.9 billion to $4.2 billion.

Those figures do not establish how much of future revenue will come from Goldman's revised 67 GW market. They do show that BE's data-center power exposure is already commercial rather than purely thematic.

That makes BE the cleaner behind-the-meter business proxy, while GEV is the broader power-system proxy.

The options markets reverse part of the company-size intuition#

The September 23 options snapshot creates a useful cross-market tension.

GEV was at $950.81 with aggregate ATM implied volatility of 46.1%, about 214,400 contracts of total open interest and 12,700 contracts of session volume. The reported average bid-ask spread across the chain was 13.96%.

BE was at $274.41 with ATM implied volatility of 75.6%, about 888,600 contracts of total open interest and 86,700 contracts of session volume. Its average bid-ask spread was 6.17%.

The smaller company therefore had roughly four times as much listed-options open interest and a much tighter reported average spread in the same vendor snapshot.

Absolute implied volatility tells the opposite story. BE was carrying far more expected price variability than GEV.

That difference should not be interpreted as the market assigning a larger Goldman-specific event premium to Bloom. The two companies have very different business concentration, equity histories, betas and company-specific risks.

The useful observation is structural: the more concentrated behind-the-meter proxy also has the deeper and mechanically cleaner listed-options market, but it carries much higher absolute volatility.

High absolute IV does not mean BE is historically expensive#

A same-underlying baseline changes the reading again.

On September 22, BE's ATM implied volatility was 76.5%, while its 20-day realized volatility was 72.4% and its one-year IV rank was reported at 0.0%. Earlier in September, BE ATM IV had been as high as 89.6%.

GEV's September 23 ATM IV was 46.1% against 48.8% 20-day realized volatility, with a one-year IV rank of 13.9%.

Both options markets were therefore in the lower part of their own recent implied-volatility histories even though BE's absolute IV was much higher than GEV's.

This is why comparing 75.6% with 46.1% alone is incomplete. Cross-company IV mainly says the market expects BE's shares to move more. The own-history comparison asks a different question: whether the current uncertainty is unusual for that company.

On the available data, the Goldman forecast revision does not coincide with historically extreme implied volatility in either underlying.

Long-dated options fit the forecast better than a one-day event study#

The catalyst itself is structural rather than scheduled.

Goldman's capacity, power-demand and behind-the-meter forecasts run to 2030. There is no single earnings-like date on which 67 GW either happens or fails.

That changes the useful options horizon.

GEV's term structure in the September 23 snapshot rises from 46.1% around October expirations to roughly 51-52% in the longest 2028 and January 2029 contracts. BE rises from the mid-70s in October to roughly 84% in late-2028 and January 2029 expirations.

The upward slopes cannot be attributed to data-center power demand. Long-dated contracts contain several years of company, technology, capital-spending, commodity and macro uncertainty.

They do, however, show why the research question belongs farther out on the curve than a one-week headline reaction. The relevant evidence will arrive through order growth, manufacturing capacity, project deployments and changes in the mix between grid and onsite power.

A September 23 same-day snapshot also cannot establish a clean pre-versus-post response to the Goldman research because the exact publication time and market-observation sequencing are not sufficiently precise.

The real options question is concentration versus diversification#

The Goldman revision creates a better framework than a generic data-center-power story.

GEV has multiple routes into the forecast. Gas turbines can serve behind-the-meter demand, while Electrification can benefit from data-center grid investment and eventual interconnection. That diversification makes the business exposure broader than the specific 67 GW onsite forecast.

BE has a narrower transmission channel. Fuel cells sit directly inside onsite generation, and the Oracle agreement demonstrates that multi-gigawatt AI deployments are already moving from concept to contracted capacity. The concentration makes the underlying economically easier to connect to behind-the-meter adoption, but it also leaves the company more sensitive to fuel-cell execution, manufacturing scale and technology-specific competition.

The options markets add another layer. BE currently combines higher absolute volatility with greater listed open interest and tighter chain-wide spreads. GEV combines lower absolute volatility with a broader business exposure and a thinner options market by those measures.

Neither structure establishes which company will capture more of the 67 GW forecast.

What would make the options comparison more informative is operating evidence that separates the power architectures: new multi-gigawatt fuel-cell deployments, changes in gas-turbine reservation volumes, data-center electrification orders, manufacturing expansions, or disclosures showing how much capacity is moving behind the meter rather than waiting for grid connection.

Those observations can then be matched to changes in BE and GEV term structures using the same data methodology.

Goldman's 217 GW headline says the data-center buildout is getting larger. The 67 GW behind-the-meter revision says something more useful for options research: the way that buildout is powered may be changing too. GEV and BE expose different pieces of that transition, and the divergence between their business concentration, volatility and liquidity is the part worth measuring.