Written by: Rita
Tide Guide
Data centers are facing an electricity crisis. GPUs can ramp up from 0% to 100% in just milliseconds, overwhelming the grid. A storage report released by Goldman Sachs on July 16 concluded straightforwardly: expanding traditional power grids takes four to eight years, while energy storage can be deployed in twelve to eighteen months—making it the fastest way for data centers to bypass the queue.
Goldman Sachs’ utilities team has just raised its forecast for U.S. electricity demand growth from 2.6% to 3.2%, with data centers being the largest variable. Of this increase, 0.5 percentage points stem from the “behind-the-meter” model in which data centers generate and store their own power, equivalent to approximately 3 GW of DC electricity demand annually by 2030.
Goldman Sachs estimates that the U.S. market alone will see an additional 50 GWh of behind-the-meter energy storage by 2030. The global energy storage market is projected to reach 2,100 GWh in annual installations by 2040. From CATL to Tesla, from FLNC to Energy Vault, energy storage companies are becoming the hidden arms dealers of AI infrastructure.
The power grid can't keep up with the speed of GPUs.
The power consumption logic of data centers has changed. Previously at 5 to 10 megawatts, they now range from 100 to 200 megawatts, with large campuses reaching gigawatt levels. Rack density has surged from 5 to 10 kilowatts to over 50 to 100 kilowatts. Even more critical, GPU loads can spike from 0% to 100% within milliseconds, creating massive strain on the power grid.
Gas turbines are not designed for this task. Industry feedback indicates that gas turbines can only start and stop twice per day, as frequent switching accelerates wear. There have been cases where new combined-cycle gas turbines experienced shaft failures after just seven months due to frequent load fluctuations.
Energy storage can solve this problem. With response times in milliseconds and deployment cycles of 12 to 18 months, it is significantly faster than grid expansion. Bypassing congested grid connection queues is energy storage’s greatest competitive advantage. Goldman Sachs is defining it as an independent business model, not merely as a backup power concept.
Where does the 50 GWh increase come from?
Goldman Sachs Asia Battery Team estimates that behind-the-meter storage opportunities in the U.S. will contribute approximately 50 GWh by 2030, plus an additional 11 GWh from 800V DC data centers, bringing the total U.S. energy storage deployment to 172 GWh—a significant increase from the previous estimate of 112 GWh.
Goldman Sachs’ utilities team has just raised its forecast for U.S. electricity demand growth from 2.6% to 3.2%, with 0.5 percentage points attributed to behind-the-meter demand, equivalent to approximately 3 GW of DC power demand annually by 2030. This is a structural shift, not a marginal change.
Globally, Goldman Sachs expects annual global energy storage installations to reach approximately 2,100 GWh by 2040. Energy storage demand from data centers is becoming a key driver of this growth curve. Forecasts for the EMEA region have been raised from 83 GWh to 142 GWh, led by Germany, Spain, and Portugal. Residential energy storage in Australia is also doubling.
Energy storage companies are becoming arms dealers for AI infrastructure.
Goldman Sachs detailed companies worth noting in the energy storage supply chain in its report:
FLNC (Buy, Target Price $22) has secured exclusive battery partnership status for NVIDIA’s DSX Vera Rubin reference architecture. It has signed master supply agreements with two hyperscale customers and is currently bidding on a 12 GWh order. As of Q2 FY2026, its data center pipeline projects total 12 GW, up 30% quarter-over-quarter. Goldman Sachs estimates this business could contribute approximately 8% to 14% of revenue by 2028.
CATL (Buy) holds approximately 30% of the global energy storage battery market and has explicitly identified data centers as a core application scenario. The TENER 6.25MWh system has been deployed at Shanghai's SenseTime AI data center, reportedly reducing electricity costs by 7%. Sodium-ion energy storage is expected to begin delivery by the end of 2026.
Tesla (neutral) set a record with 46.7 GWh of energy storage deployments in 2025, with even higher volumes expected in 2026. Megapack and Megablock are becoming critical grid-stabilizing infrastructure for hyperscale data centers. Goldman Sachs forecasts Tesla’s energy business to generate $29 billion in revenue by 2028, accounting for approximately 21% of the company’s total revenue.
Energy Vault (neutral, target price $2.50) is transitioning from an energy storage system provider to an energy infrastructure platform. Goldman Sachs has included the Asset Vault business in its valuation for the first time, applying a 6x EV/EBITDA multiple. The business is expected to achieve an EBITDA margin of approximately 64% by 2028.
LGES (Buy) is converting five North American manufacturing sites into dedicated ESS production lines, with North American capacity expected to reach 50 GWh by the end of 2026. The share of energy storage revenue is projected to rise from approximately 14% in 2025 to around 31% in 2026. Through its subsidiary Vertex, LGES offers an integrated “one-stop” solution combining batteries and software, bundled with the group’s proprietary liquid cooling and data center design services.
Other notable targets: Canadian Solar (Sell, subsidiary e-STORAGE has 81 GWh of project backlog); Ford (Hold, Ford Energy expected to generate $3 billion in revenue by 2028); Samsung SDI (Hold, ESS battery revenue share increased to approximately 26%); Shoals (Buy, BESS product revenue expected to reach $30–35 million in 2026); Sungrow (Hold, energy storage contributes approximately 42% of revenue).
Tide perspective
The most significant contribution of Goldman Sachs' report is transforming "data center energy storage" from a concept into a quantifiable market. A 50 GWh increase in behind-the-meter storage means that energy storage is evolving from a supplementary component for renewable energy into a critical necessity for AI infrastructure.
This shift in identity changes the entire industry's valuation logic. Energy storage companies are no longer solely tied to the pace of solar and wind power; they now directly benefit from the expansion of AI capital expenditures. Goldman Sachs raised FLNC’s target price from $20 to $22, based on increasing visibility of data center orders.
But this report also needs to be read in reverse. The report from Goldman Sachs includes targets rated as Buy, Neutral, and Sell; investors need to distinguish which ones are truly supported by orders and which are merely tangentially related to the theme. FLNC has an exclusive agreement with NVIDIA, CATL has actual shipments, and Tesla has deployment data. The gap between these companies and those merely “developing related products” is real.
Disclaimer
This article is a compilation and interpretation by Chaoxiang Research of a third-party brokerage research report (Goldman Sachs, July 16, 2026). The ratings, price targets, earnings forecasts, and related judgments cited herein reflect the views of the brokerage's analysts and represent the position of their respective institution only; they do not reflect the views of Chaoxiang Research nor constitute any investment advice.
The market carries risks; make decisions independently. This article should not be used as a basis for buying or selling any securities.