California's Battery Storage Fleet Is Leaving $98 Million on the Table: Inside the 2.4× Revenue Optimization Gap — Analysis
A landmark study by electricity optimization company Gridmatic has quantified what many in the energy storage industry have long suspected but could never prove: California's grid-scale battery storage fleet is dramatically underperforming relative to its theoretical revenue potential, and the gap is driven by operational strategy choices — not technology limitations. The "CAISO Energy Storage Report 2024-2025," produced using Gridmatic's proprietary shadow clearing engine that reconstructs what each of 30 large BESS assets across the California Independent System Operator market could have earned with optimal dispatch, reveals a performance chasm. Monthly merchant revenue (energy plus ancillary services) ranges from below US$1/kW-month to above US$6/kW-month, with the median at just US$2.67/kW-month. Only five of the 30 studied assets outperformed the theoretical "TBx" optimal benchmark, and the fleet's median capture rate was just 82.3%. Gridmatic calculates that if all 30 assets achieved the optimization level of the top performer — the 100MW/400MWh Caballero BESS — the fleet would have generated an additional US$98 million in revenue over the two-year study period. This analysis examines what the data reveals, why static bidding strategies are the primary culprit, and what the findings mean for storage operators, investors, and the rapidly growing CAISO storage fleet.
Overview of Gridmatic's Shadow Clearing Methodology and the 30-Asset CAISO Fleet
Gridmatic's study is methodologically notable because it does not rely on survey data, operator self-reporting, or aggregated market statistics. Instead, the company used its proprietary shadow clearing engine — a full simulation of CAISO's day-ahead and real-time market clearing processes — to reconstruct the hourly dispatch and revenue outcomes for each of the 30 studied assets. This approach addresses the fundamental challenge of BESS benchmarking: unlike a natural gas plant whose optimal dispatch is straightforward (run when the spark spread is positive), a battery's optimal dispatch across energy and multiple ancillary service products is path-dependent — today's dispatch decision constrains tomorrow's state-of-charge, making "what could have been earned" a complex counterfactual calculation rather than a simple market-price lookup.
The 30 assets in the study represent approximately 2.3GW of total power capacity — roughly 25-30% of CAISO's grid-scale BESS fleet at the time of the study — and span a range of sizes, vintages, locations, and commercial structures. The sample includes utility-owned assets dispatched by in-house trading desks, assets operated under tolling agreements by third-party optimizers, and merchant assets bidding directly into CAISO markets. This diversity makes the study's central finding — that performance variation is enormous and systematically linked to bidding strategy — all the more significant, as it suggests that strategy trumps structure, and that even assets with suboptimal commercial arrangements could dramatically improve performance through better operational decision-making. Discover AGAIC POWER's grid-scale battery storage systems for electricity market optimization.
Why the CAISO Performance Gap Exists — and Why It Persists
The raw numbers are striking: monthly merchant revenue ranging from under US$1 to over US$6 per kilowatt-month — a 6× spread that cannot be explained by differences in asset size, location, duration, or age. The median revenue of US$2.67/kW-month and the 82.3% median TBx capture rate together suggest that the typical CAISO BESS asset is capturing approximately four-fifths of what it could earn with optimal dispatch — a gap that translates to roughly US$0.58/kW-month of foregone revenue at the median, or approximately US$700,000 per year for a 100MW asset.
Gridmatic's diagnosis of the root cause is devastatingly specific: bidding strategy rigidity. Twenty of the 30 studied assets used fixed frequency regulation capacity curves in more than 75% of study months — meaning they allocated a pre-determined portion of their capacity to frequency regulation (regulation up and regulation down) regardless of whether that day's market conditions favored regulation, spinning reserves, or energy arbitrage. Thirteen of the 30 assets used static energy bids — meaning they submitted the same price-quantity pairs regardless of time of day, season, or market conditions — for more than half of all study hours. These are not sophisticated optimization failures; they are process failures — the result of a BESS being treated as a "set and forget" asset rather than an actively managed trading resource.
The persistence of these static strategies, despite the demonstrable revenue penalty, points to a structural issue in the BESS operations market. Many early BESS installations were procured as reliability assets — contracted to provide resource adequacy capacity or frequency regulation — with merchant revenue treated as ancillary income rather than the primary value driver. The commercial agreements, operational protocols, and trading desk capabilities established for these reliability-focused assets are ill-suited to the dynamic optimization that merchant revenue maximization requires. Changing these structures — renegotiating tolling agreements, upgrading SCADA and market interface systems, building or contracting algorithmic trading capability — involves transaction costs and organizational friction that many asset owners have been slow to overcome, even when the revenue upside is substantial.
Technical Deep Dive: Why Static Bidding Underperforms Dynamic Optimization in Multi-Product Markets
The technical explanation for why static bidding strategies fail in CAISO lies in the market's product structure. CAISO operates separate day-ahead and real-time markets for energy, four regulation products (regulation up/down, each with mileage and capacity components), and multiple contingency reserve products (spinning and non-spinning reserves). Each product has its own hourly clearing price, and a BESS's capacity allocated to one product cannot simultaneously be offered into others — creating an opportunity cost calculation that changes every hour based on relative product prices, forecasted net load, renewable generation, and system conditions.
A static bidding strategy — for example, always offering 20MW of regulation up and 20MW of regulation down while leaving the remaining capacity for energy arbitrage — makes the implicit assumption that the average revenue from this fixed allocation, across all hours of all days, exceeds what could be earned from any alternative allocation. This assumption is provably false in markets with significant price volatility. On a high-solar day with steep midday-to-evening price ramps, the energy arbitrage opportunity may be worth US$200-300/MWh for a 4-hour discharge, while frequency regulation capacity may clear at only US$5-10/MW-hour. On a low-load, high-wind night, frequency regulation prices may spike to US$50-100/MW-hour while energy spreads flatten to near zero. The optimal allocation is fundamentally time-varying, and a static strategy guarantees that the asset is misallocated in both extremes.
The path-dependency of BESS optimization compounds this problem. A static strategy that allocates 40MW to regulation in a given hour not only forgoes the energy arbitrage opportunity in that hour but also constrains the state-of-charge available for subsequent hours. If the battery provides 40MW of regulation down (charging) during an hour when energy prices are already low, it may reach full state-of-charge and be unable to charge further during even lower-price hours later in the day — missing the cheapest energy purchase opportunity. Dynamic optimization frameworks solve this by treating the BESS as a stochastic dynamic programming problem: at each decision interval, the operator evaluates not just the immediate revenue from each allocation option but the expected value of the state-of-charge position that each option creates for future intervals. This requires forecasting future price distributions — not just point estimates — across multiple products, a computational task that is infeasible without specialized optimization software. Explore our energy storage solutions with integrated market optimization capabilities for CAISO and other wholesale markets.
Real-World Applications: The Caballero Case Study — 141% TBx Capture
The study's standout performer is the Caballero BESS — a 100MW/400MWh asset owned by Alpha Omega Power and optimized by Gridmatic. Caballero achieved a 141% TBx capture rate, meaning it earned 41% more than the "theoretical best" benchmark. This apparently impossible result — how can an asset earn more than the theoretical maximum? — is explained by the TBx benchmark's design limitations: TBx assumes perfect foresight of market prices but does not account for physical constraints and market rules that a well-designed but imperfect optimization can exploit. In practice, a sophisticated optimizer that understands CAISO's market rules, bidding deadlines, and physical asset constraints can exploit structural inefficiencies in market design — such as the temporal gap between day-ahead market closure and real-time dispatch, during which updated forecasts can reveal profitable re-dispatch opportunities — that the TBx benchmark's simplified assumptions miss.
Caballero's median revenue of approximately US$6.55/kW-month (at 141% capture of the fleet median of US$2.67/kW-month) represents roughly 2.45× the fleet median — translating to approximately US$7.9 million in additional annual revenue for a 100MW asset compared to the typical CAISO BESS. This performance premium is not the result of a one-time strategy optimization but of continuous algorithmic trading: Gridmatic's system re-optimizes Caballero's dispatch every five minutes based on updated market data, weather forecasts, and asset state. The contrast with the fleet's median 82.3% capture rate illustrates that BESS optimization is not a technology problem that can be solved once and forgotten — it is an ongoing operational process where sustained investment in people, software, and market intelligence directly translates into revenue outperformance.
Industry Impact: What the CAISO Gap Means for the Global Storage Market
The Gridmatic study's implications extend far beyond California. CAISO is the largest and most liquid BESS market in the United States, with approximately 10GW of operational and in-construction storage capacity. If the US$98 million two-year revenue gap for the studied 2.3GW sample scales to the full fleet, the CAISO-wide revenue optimization gap could exceed US$400 million annually — a number that will grow as the state's storage capacity expands toward the California Public Utilities Commission's 52GW by 2045 planning target.
The operational lessons are equally transferable to other organized wholesale markets — ERCOT (Texas), PJM, MISO, NYISO, and ISO-NE all operate multi-product markets with structural similarities to CAISO where BESS assets face analogous optimization challenges. The static bidding patterns that Gridmatic documented in California — fixed frequency regulation curves, fixed energy bids, infrequent strategy updates — are likely present in other markets where BESS assets were procured as reliability resources rather than merchant trading assets. The revenue uplift opportunity from dynamic optimization is proportional to market price volatility and product complexity — attributes that characterize all major US wholesale markets and are increasing as renewable penetration grows.
A secondary but important finding concerns asset availability. The study found a fleet-wide average availability of just 79.7% — meaning that on average, BESS assets were unavailable for dispatch during 20.3% of hours. The Gateway project's 24-month outage following a fire event and Vistar Moss Landing's continued offline status since the January 2025 fire demonstrate that availability risk in BESS is concentrated in catastrophic failure events rather than distributed across routine maintenance downtime. For a merchant BESS, every hour of unavailability is an hour of foregone revenue — and at a median of US$2.67/kW-month, a 100MW asset losing 20% availability is foregoing approximately US$640,000 annually in merchant revenue alone, not including the reputational and contractual costs associated with reliability events.
Future Outlook: The Optimization Software Arms Race
The CAISO data makes an unambiguous case for investment in BESS optimization capability, and the market is responding. Independent software platforms — Gridmatic, Habitat Energy, Tesla Autobidder, Stem's Athena, and others — are competing to provide algorithmic trading and dispatch optimization for BESS assets. The value proposition is straightforward: a platform capable of improving capture rate from 82.3% to 110% (still below Caballero's 141%) would generate approximately US$3.3 million in additional annual revenue for a 100MW asset at current CAISO prices — an order of magnitude more than the software licensing and operations cost.
The broader industry implication is that BESS is transitioning from a hardware business — where competitive advantage came from lower battery cell costs and better system integration — to a services business, where competitive advantage comes from software, market intelligence, and operational expertise. This mirrors the evolution of other capital-intensive industries: airlines migrated from competing on aircraft ownership to competing on revenue management, and telecommunications migrated from competing on network infrastructure to competing on service platforms. For storage investors, the lesson of Gridmatic's study is clear: the quality of the optimization platform may be a larger determinant of BESS investment returns than the cost of the underlying battery hardware.
