Primary-Source Investigations
The Megapack Maneuver
Tesla, Inc., Form 10-K, fiscal year ended December 31, 2020. EDGAR Accession No. 0001564590-21-004599. Filed February 8, 2021.
I am reading a filing published by Tesla, Inc. on February 8, 2021, that reports automotive regulatory credits of $1,580 million, and I am telling you that this line is not a footnote. It is the architecture.
The line sits in Item 7. Management's Discussion and Analysis. Four paragraphs into Automotive Revenue, set in the same Times New Roman as every other disclosure — one line. Automotive regulatory credits: $1,580 million. No exclamation point. No asterisk. One billion five hundred and eighty million dollars. For that fiscal year, the difference between Tesla's first full-year GAAP net income and another loss.
It is a filed document. It is public. It is dated.
I know how this works because I built the system that tracks how this works. TELOS. The pipeline. The substrate. I built it because I got tired of reading "widely reported" and "many believe" in every article about electric vehicles and subsidies, and I wanted a machine that would only accept claims with a filing number and an archive location. The machine does not care about my opinion. The machine only cares whether the source is named and filed. But I am the operator, and I am sitting here at 3:47 AM reading Tesla's Form 10-K, and I am telling you that the $1,580 million line is not revenue from selling cars. It is a transfer mandated by state law, collected from other automakers, and booked as revenue by the company that over-complied with the mandate. The mandate is the architecture. The architecture is filed.
The California Air Resources Board created the unit. Title 13 of the California Code of Regulations, § 1962.2 — the Zero-Emission Vehicle Regulation. The rule: every manufacturer selling above a threshold into the California market must either produce a certain percentage of zero-emission vehicles or buy credits from a manufacturer who over-produced. The Board sets the arithmetic. The Board conducts the audit. The Board issues the credits. The Board records the transfers. The Board operates out of 1001 I Street, Sacramento, California 95814, in a building paid for with public funds, on a budget approved by the California Legislature.
The California rule does not stand alone. Section 177 of the federal Clean Air Act, 42 U.S.C. § 7507, allows any state to adopt the California vehicle-emission standards in lieu of the federal standards, provided the adoption is identical and follows the same lead-time. New York adopts at 6 N.Y.C.R.R. Part 218. New Jersey adopts at N.J.A.C. 7:27-29 and 7:27-31. Massachusetts adopts at 310 C.M.R. § 7.40. Colorado adopts at 5 C.C.R. § 1001-24, Regulation No. 20, phasing in for model year 2023. Oregon adopts at OAR 340-257-0010 through 340-257-0050. Washington State adopts under RCW 70A.30 and WAC 173-423. Minnesota adopts under Minn. R. ch. 7023, the Clean Cars Minnesota Rule, effective model year 2025.
Seven Section 177 states beyond California. Seven statutory references in seven administrative codes. Seven effective dates across three decades. Seven under-complier counts feeding one consolidated credit market the over-complier — Tesla — sells into. One architecture. Several signatories. One line on Item 7.
Here is how the credit business runs. Read it once and the geometry locks in.
You build an electric vehicle factory. You over-produce against the California mandate. You take delivery of credits issued by the Board. The credits are not cash. The credits are an instrument. The instrument has a counterparty. The counterparty is every manufacturer in the state market that missed the mandate. General Motors. Stellantis. Ford. Toyota in the years Toyota under-complies. The counterparty has two choices. Pay the Board's fine. Pay you. The fine is set so the second choice is always cheaper. You name your price below the fine and above your cost basis. You sign a credit-transfer agreement. Bilateral. Private. Logged with the Board for audit, not published line-by-line with the buyer's name and the price. The wire clears. The credit transfers. The Board updates the ledger. You book the revenue. You disclose the aggregate on Item 7. You do not disclose the buyer-by-buyer breakdown because the SEC does not require it.
That is the entire business. That is the line on the 10-K. One billion five hundred and eighty million dollars in fiscal year 2020. No cost of goods sold. No warranty reserve. No inventory carry. The residual of having built a car factory that met a number the state set in advance. The state set the number to force the build. The build was the policy goal. The residual is the cash. The cash subsidizes the next factory. The next factory builds more cars. The cars accumulate more credits. The credits sell to the same counterparties for another year. The flywheel turns. The Board is the bookkeeper. The counterparty is the bank. The state is the architect. The architecture is filed.
The other filing lands two segments away. Tesla, Inc., Quarterly Report on Form 10-Q for the quarter ending 2024. Same CIK. Same registrant. Different segment: Energy Generation and Storage. Line item: Megapack revenue. The Megapack is a utility-scale lithium iron phosphate battery storage system, manufactured at the Lathrop, California Megafactory, intended for grid operators and large commercial customers. The 10-Q discloses deployments at per-quarter granularity. Numbers in gigawatt-hours. Revenue in dollars. Growth rate in percent.
The Lathrop Megafactory is a single building on Roth Road, on a parcel of agricultural-into-industrial conversion Tesla acquired in tranches between 2021 and 2023. The building is engineered for one product: the Megapack. The Megapack ships pre-integrated, pre-tested, pre-certified, on a flatbed truck or a shipping container, to a customer site, where it is connected, configured, and energized. Days from factory to energization. The transformer the same customer would need from a conventional utility supply chain takes one hundred and twenty weeks. The Wood Mackenzie utility-transformer market report, 2024 edition, documents the lead-time blowout. The EIA Annual Energy Outlook 2024 documents the demand surge — Table A2 of the reference case shows commercial electricity demand projected to surpass residential by 2027.
One hundred and twenty weeks. Twenty-eight months. Two and a third years. The developer cannot wait. The developer has hyperscaler contracts. The hyperscaler has model-training schedules. The schedule is the only schedule that matters. The Megapack ships in days. The transformer arrives in years. The choice is a non-choice. The choice is the architecture.
Here is how the Megapack sale closes. You are a developer in Phoenix or Mesa or Goodyear. You have signed a lease with a hyperscaler. They want forty megawatts of compute online in six months. You apply to Arizona Public Service or Salt River Project for the interconnection upgrade. The utility tells you the interconnection timeline is the transformer plus the substation upgrade plus the queue position. The queue is whatever it is. The transformer is one hundred and twenty weeks. The substation upgrade is another twelve. You add the numbers. You miss the deadline by a year. You lose the lease. You pay the penalty. You go bankrupt.
You do not go bankrupt. You call Tesla. The account team is in your office in a week. They quote a Megapack array sized to your forty megawatts plus inverter losses plus redundancy headroom. The quote includes unit count, integration timeline, delivery schedule, commissioning crew, warranty terms. Email three days after the site visit. You sign. You wire the deposit. Lathrop schedules the build into a slot four weeks out. Trucks roll in week seven. Crew arrives in week eight. System energizes in week twelve. You keep the lease. You ship the workload. The hyperscaler trains the model. The model is announced from a stage in Mountain View or Menlo Park or Cupertino. The model runs on the workload that runs on the racks that run on the Megapack that shipped from Lathrop, whose bills clear out of the same Energy Generation and Storage segment the 10-Q discloses one line below the Automotive Regulatory Credits line that disclosed one billion five hundred and eighty million dollars two years before any of this happened.
That is the close. That is the maneuver. That is the cost stack.
The maneuver repeats. You are in Loudoun County, Virginia, on the Route 28 corridor between Sterling and Ashburn, on a parcel the county's Department of Economic Development inventories under data-center industrial. The corridor carries the largest concentration of data-center capacity in the world by megawatts of installed compute load. Electric service comes from Dominion Energy Virginia. The queue for a forty-megawatt load on the corridor is whatever the queue says it is. The transformer the utility would need is the same one-hundred-and-twenty-week unit the Wood Mackenzie 2024 report documents. You call Tesla. The account team is on the corridor in a week. They quote the array. You sign. You wire. Trucks roll. The array energizes in week twelve. Your Dominion cost-of-service line stays where it is. Your residential neighbors' cost-of-service line stays where it is. The transformer queue stays where it is. The Tesla Megapack ledger picks up the new array.
Here is how the cost flows. You live in Phoenix. You pay an electric bill to Arizona Public Service. APS owes a return on equity to shareholders under the rate case it filed with the Arizona Corporation Commission. The rate case includes a cost-of-service allocation that distributes grid-investment costs across the customer base. The grid-investment costs include the transformer procurement budget. The procurement budget is gross because the transformer market is supply-constrained. The constraint pushes the unit price of every transformer the utility buys. The unit price flows through the rate case. The rate case sets your bill. You pay the bill.
The data center down the road does not wait one hundred and twenty weeks. The data center buys the Megapack. The data center bypasses the utility-side transformer for inverter-tied energy storage. Load comes online in months instead of years. The data center bills the hyperscaler for compute. The hyperscaler bills its customers for inference and training. The customers are the public. The public pays for inference and training in the price of every consumer product with a large language model wedged into its support stack. The public also pays for the utility-side transformer through the rate base. The public also pays for the California regulatory credit through the price of every internal-combustion vehicle in the under-complier's product line. The public pays at the front. The public pays at the back. The public pays in the middle. The only entity in the chain that has structurally arbitraged ratepayer payment against credit-buyer payment is the one entity that built the factory in time.
Walk the year-over-year. Read it slow. The same line — Automotive regulatory credits — appears on Item 7 of every Tesla annual report from the first profitable credit-sale year forward. Each annual filing carries a different accession number. Each accession resolves to a PDF on EDGAR. Each PDF runs between two hundred and three hundred pages. Each PDF sets the credit line in the same Times New Roman as the rest of the disclosure. Fiscal year 2018. Fiscal year 2019. Fiscal year 2020 — one billion five hundred and eighty million dollars. Fiscal year 2021. Fiscal year 2022. Fiscal year 2023. Fiscal year 2024. Seven fiscal years. Seven accession numbers. Seven Item 7 line items. One ledger row in Sacramento underneath each. One ZEV mandate underneath the ledger.
The under-complier list shifts year to year. The mandate does not. The mandate is in Title 13. At § 1962.2. Signed.
The regulatory credit revenue is one number. The Megapack revenue is another number. The transformer queue is the constraint that makes the second number possible. The state policy is the architecture that made the first number possible. Two numbers, one company, one regulator, one portal. The portal is public. The filing is public. The numbers have always been public. The architecture has always been public. The reader who pulled the 10-K in February 2021 saw the number. The reader who pulls it in May 2026 sees the same number. The number has not moved. The architecture has not moved. The state policy has not moved. The only thing that has moved is the consequence — the public ratepayer underwriting the transformer queue that funnels the data center buyer into the Megapack contract that runs the hyperscaler workload that trains the model that prices the product the public buys back at a markup that includes the cost of the workload that runs on the rack that runs on the Megapack that shipped from the factory that pays its electric bill out of the same balance sheet that booked one billion five hundred and eighty million dollars in residual from a ZEV credit issued by a state agency in Sacramento.
The consequence has moved. The filings have not. The filings were where they were filed.
EDGAR Accession No. 0001564590-21-004599. Tesla, Inc., Form 10-K, fiscal year ended December 31, 2020. Item 7. Automotive regulatory credits: one billion five hundred and eighty million dollars. Filed February 8, 2021. Public.
The system is not broken. The system is working exactly as designed.
Primary sources: SEC EDGAR (Tesla 10-K FY2020 Accession 0001564590-21-004599); California Air Resources Board ZEV program records (Title 13 C.C.R. § 1962.2; Cal. Health & Safety Code §§ 43018, 43018.5); EIA Annual Energy Outlook 2024 (Tables A2, A8); Wood Mackenzie utility-transformer market report 2024.