SpaceX and Tesla’s joint chip venture, Terafab, is now targeting the third quarter of 2028 to begin volume chip production, Elon Musk said during a SpaceX conference call reported by Reuters on October 20, 2026. Musk also floated a wider window, saying the fourth quarter of 2028 remains on the table, and he was careful to frame the date as a target rather than a promise.
The disclosure is the clearest public timeline yet for a project that Musk has pitched as critical to both companies’ AI and robotics ambitions. Terafab is meant to supply chips for Tesla’s Optimus humanoid robots and self-driving Cybercabs, while also producing higher-power silicon for SpaceX’s space-based data center plans. Reuters, Space.com, and the South China Morning Post all covered the remarks, a story also syndicated by regional outlets including the Modesto Bee, underscoring how closely investors are tracking Musk’s chip manufacturing bet alongside his rocket and electric-vehicle businesses.
What Is Terafab? SpaceX and Tesla’s Joint Chip Factory
Terafab is the name Musk has given to a chip manufacturing effort shared between SpaceX and Tesla. Rather than relying solely on outside foundries, the two companies are working toward building their own capacity for advanced semiconductors, according to the announcement covered by Reuters. The project’s long-term production objective, as described by Musk, is roughly 1 terawatt per year of computing power dedicated to AI and robotics workloads.
That figure is the clearest public signal of how large Musk wants Terafab to become. A terawatt a year of compute is a scale normally associated with the biggest cloud and AI infrastructure buildouts, not a single company’s internal chip supply. Musk’s own framing of the plan, as relayed by Reuters, acknowledges how ambitious the schedule is: “So these are obviously like insanely fast time frames for chip production,” he said on the call.
Unlike a conventional foundry announcement with a groundbreaking date and a ribbon-cutting, Terafab has so far been described mostly through Musk’s public remarks rather than a formal corporate filing or press release from either SpaceX or Tesla. That makes the October 20 call, and the named outlets that reported on it, the primary record of what the companies are actually targeting.
Musk’s Q3 2028 Target, and the Error Bars Around It
The headline figure from the call is simple: Terafab aims to begin volume production in the third quarter of 2028. But Musk immediately qualified that number rather than presenting it as a locked-in commitment. “I put big error bars on this because we don’t know if we will actually achieve this or not,” he told the call, according to Reuters.
He then widened the window further, noting that the fourth quarter of 2028 is also plausible. “Let’s just say that outcome of achieving volume production is at least in the set of possible outcomes for Q3 or Q4 of ’28,” Musk said. That phrasing, choosing to describe an outcome as merely “in the set of possible outcomes” rather than guaranteed, is characteristic of how Musk has talked about ambitious manufacturing timelines in the past, and it is worth taking at face value here rather than rounding it up to a firm promise.
For a company with Tesla’s track record on vehicle production ramps and Model lines, and SpaceX’s history of slipping its own launch and development schedules before eventually delivering, the hedge is notable. Musk is on record setting aggressive dates for Starship test flights and Full Self-Driving milestones that moved multiple times before being hit, or in some cases are still pending. The Terafab timeline looks built from the same playbook: an aggressive headline number paired with an explicit acknowledgment that it might not happen on schedule.
The First Test Chip: An SRAM Milestone in Austin
Before any talk of volume production, the companies have a nearer-term checkpoint. Musk said Terafab’s research facility in Austin, Texas is targeting a functional SRAM (static random-access memory) test chip around May or June of 2027, per the reporting reviewed from the call. That date sits more than a year ahead of the Q3 2028 volume-production goal, effectively serving as the proof-of-concept gate the rest of the schedule depends on.
SRAM is a logical starting point for a new fab’s test chip. It is one of the more straightforward circuit types to manufacture and verify, which makes it a common choice for shaking out a new process before committing to the far more complex logic chips that would eventually power products like Optimus or Cybercab. If the Austin facility hits that mid-2027 SRAM milestone, it would be the first concrete, verifiable signal that Terafab’s manufacturing process is working at all, ahead of the harder scale-up to commercial volumes.
Beyond the SRAM test chip and the Austin location, the specifics remain limited. The exact manufacturing process node, wafer-start capacity, projected annual chip output, facility construction cost, and how production will be split between SpaceX’s and Tesla’s needs have not been detailed in the reporting around this call. Readers should treat any more specific figures circulating elsewhere with caution until SpaceX or Tesla confirm them directly.
Why Tesla and SpaceX Want Their Own Chips
The strategic logic behind Terafab is straightforward even if the execution is not: both companies have businesses that are increasingly bottlenecked by access to advanced chips, and neither wants to depend entirely on third-party foundries for parts central to their roadmaps. Tesla’s push into humanoid robotics with Optimus and its ongoing self-driving ambitions for Cybercab both require large volumes of efficient edge-computing and inference silicon. SpaceX, meanwhile, has been expanding into AI infrastructure and has talked about space-based data centers that would need high-power chips able to survive and perform in orbit.
That dual demand is a big part of why Musk has framed Terafab’s scale in terawatts rather than simple chip counts. Optimus robots at any meaningful production volume would need a steady, high-reliability supply of inference chips. Cybercab’s self-driving stack adds another layer of edge-compute demand. And SpaceX’s orbital data center ambitions point toward chips that need to handle thermal and power constraints unlike anything running in a terrestrial data center. Tesla has separately faced reported friction with outside chip partners over its AI5 and AI6 silicon, a dynamic that adds extra weight to the case for bringing more chip production in-house.
Owning more of the chip supply chain also insulates both companies from the kind of allocation squeezes that have hit the broader AI hardware market. Nvidia’s GPUs remain in high demand across the industry, and SpaceX itself has been reported to be building out a large Nvidia GPU footprint for its own AI infrastructure. Terafab, if it works, would give Musk’s companies a second lever beyond simply buying more capacity from established chipmakers.
Optimus, Cybercab, and the Edge-Computing Push
Tesla’s humanoid robot program has become one of the clearest cases for why the company might want dedicated chip capacity. Optimus units need onboard inference silicon that is efficient enough to run on a mobile power budget while still handling complex vision and motor-control tasks. The faster and cheaper Tesla can source that chip, the more it can control costs as it scales the robot program toward higher weekly production numbers.
The Cybercab self-driving push has a similar logic, with the added complication that automotive-grade chips typically carry more stringent reliability and longevity requirements than consumer electronics. Building those chips in a facility designed with Tesla’s own product requirements in mind, rather than adapting an off-the-shelf part from an external foundry’s existing process, is the kind of vertical integration Musk has pursued elsewhere in Tesla’s business, from battery cells to in-house vehicle software.
None of the specific chip architectures for Optimus or Cybercab beyond this general framing have been confirmed publicly in connection with Terafab. The stated applications come from the broader reporting on the project; exact part numbers, process nodes, or performance targets for those chips have not been disclosed.
SpaceX’s Space-Based Data Center Angle
The SpaceX side of Terafab points toward a different, and in some ways more unusual, use case: high-power chips destined for space-based data centers. SpaceX has talked about using its launch cadence and satellite expertise to put computing infrastructure directly into orbit, a concept that would require chips engineered to handle the thermal extremes and radiation exposure of space rather than a climate-controlled data center floor.
That is a very different design target than the edge-inference chips Tesla needs for Optimus and Cybercab, and it is one reason Terafab’s scope, as described by Musk, spans both low-power edge silicon and higher-power parts in the same facility. Running two very different chip classes through one manufacturing buildout adds complexity, but it reflects how tightly SpaceX’s AI infrastructure ambitions and Tesla’s robotics roadmap now overlap under Musk’s combined leadership of both companies. SpaceX’s own site outlines its broader push into satellite and infrastructure manufacturing, a pattern Terafab would extend into chips.
The 1-Terawatt-a-Year Target, in Context
Musk’s long-term production objective for Terafab, a terawatt of computing power annually for AI and robotics, is the single most eye-catching figure attached to the project, and it is the detail that has drawn the most outside commentary on just how large Musk wants the facility to eventually become. It is also the hardest to independently verify this early, since it depends entirely on Terafab actually reaching the volume-production stage Musk is targeting for 2028, and then continuing to scale from there.
Put simply: hitting that kind of output would place Terafab in the same conversation as the largest chip and data center buildouts happening anywhere in the industry right now, at a time when companies across cloud computing and AI are already racing to add compute capacity. Nvidia’s own roughly $500 billion in committed AI infrastructure spending gives a sense of the capital intensity already flowing into this space, and it is the scale Terafab would need to approach relevance against.
It is worth being precise about what is and is not confirmed here. The 1-terawatt figure is Musk’s stated long-term objective, not a near-term production number, and SpaceX and Tesla have not published a breakdown of how that capacity would be split between the two companies’ needs, nor a construction cost for the facility that would support it.
Historical Context: Chip Fabs That Missed Their Own Timelines
Musk’s own hedge on the Terafab timeline lines up with a broader pattern in the chip industry: new fab construction projects routinely slip their original target dates, even when backed by established semiconductor giants with decades of manufacturing experience. Intel, TSMC, and Samsung have all announced U.S. fab expansions in recent years that moved later than their initial public timelines, often due to construction delays, equipment lead times, or shifting demand forecasts.
That track record matters for how to read Terafab’s Q3 2028 target. SpaceX and Tesla are not traditional chipmakers; neither company has previously operated a semiconductor fab at the scale Musk is describing. TSMC, by contrast, has decades of fab-building experience and still faces schedule pressure on new plants. A first-time fab operator targeting an “insanely fast” schedule, in Musk’s own words, is taking on execution risk that even the most experienced foundries have struggled with recently.
Tesla’s own manufacturing history cuts both ways on this point. The company has, at times, hit aggressive production ramps that skeptics doubted, including scaling the Model 3 and Model Y lines faster than many analysts expected. But it has also missed self-imposed deadlines repeatedly, particularly around Full Self-Driving capability timelines that Musk has revised multiple times over the past decade. Terafab’s eventual outcome will likely land somewhere in that same range of possibilities.
Terafab vs. the Established Chip Manufacturing Landscape
Terafab enters a chip manufacturing landscape already dominated by a small number of foundries with enormous scale and manufacturing experience. The table below places Terafab’s publicly stated plans alongside how the established players are generally understood to operate, based on their own public production and business models, to give a sense of the gap Terafab would need to close.
| Manufacturer | Primary Customers | Operating Model | Reported Status (as of Oct. 2026) |
|---|---|---|---|
| Terafab (SpaceX/Tesla) | Internal: Optimus, Cybercab, SpaceX AI infrastructure | Captive fab for two affiliated companies | Targeting SRAM test chip in 2027, volume production in Q3/Q4 2028, per Musk |
| TSMC | Apple, Nvidia, AMD, and other external clients | Pure-play contract foundry | Operating at scale, posting record profit amid sustained AI chip demand |
| Samsung Foundry | External clients including memory and logic customers | Integrated device manufacturer with foundry services | Supplying memory and logic chips to multiple AI hardware makers |
| Intel Foundry | External clients plus Intel’s own product lines | Integrated device manufacturer expanding foundry services | Building out U.S. fab capacity alongside its own chip lines |
| Nvidia (fabless) | Cloud providers, enterprises, SpaceX among others | Design house relying on external foundries | Committed to roughly $500 billion in AI infrastructure spending |
The comparison underscores how unusual Terafab’s structure is. Every other row in that table is either a dedicated foundry serving a wide customer base, or a fabless chip designer buying capacity from one of those foundries. Terafab, as described so far, is neither: a captive fab built by two affiliated but legally distinct companies, intended primarily to serve their own internal chip needs rather than sell capacity to outside customers. That model has precedent in the industry’s history, but rarely at the scale Musk is describing for the 2028 target.
Terafab Timeline at a Glance
Here is what has been publicly confirmed about Terafab’s schedule so far, based on Musk’s remarks on the October 20, 2026 call as reported by Reuters.
| Milestone | Target Date | Location | Confidence Level |
|---|---|---|---|
| SRAM test chip functional | Approximately May or June 2027 | Terafab research facility, Austin, Texas | Internal target, per Musk |
| Volume chip production begins | Q3 2028 (possibly Q4 2028) | Not fully specified | Explicitly described by Musk as uncertain, with “big error bars” |
| Long-term production objective | Not dated | Not fully specified | Approximately 1 terawatt per year of computing power |
What Musk Actually Said on the Call
Because so much of the public record on Terafab comes from Musk’s own remarks rather than a formal corporate disclosure, it is worth quoting him directly on the key points, as reported by Reuters from the October 20, 2026 call.
On the uncertainty of the 2028 target, Musk said: “I put big error bars on this because we don’t know if we will actually achieve this or not.”
On the window for volume production, he said: “Let’s just say that outcome of achieving volume production is at least in the set of possible outcomes for Q3 or Q4 of ’28.”
And on the overall pace of the project, Musk said: “So these are obviously like insanely fast time frames for chip production.”
Those three statements, taken together, paint a consistent picture: an executive publicly committing to an aggressive date while simultaneously warning that the date might not hold. It is a notably more cautious framing than Musk has used for some of his other public announcements, and it suggests internal awareness that first-time fab construction and ramp-up carries real execution risk, regardless of how motivated the teams building it are.
Market Impact: Why Investors Are Watching Terafab
Terafab lands at a moment when investors are already scrutinizing every piece of Musk’s sprawling business empire for signs of how SpaceX and Tesla plan to fund and staff their AI ambitions. Talk of a closer financial relationship between Tesla and SpaceX has already been circulating, and a shared chip manufacturing venture adds another layer to how intertwined the two companies’ infrastructure plans have become.
SpaceX’s own AI infrastructure bet has grown quickly in financial terms, with the company’s AI-related revenue reportedly climbing sharply over the past year. A captive chip supply through Terafab would, in theory, give SpaceX more control over the cost structure behind that growth, rather than remaining fully dependent on external GPU and chip suppliers. SpaceX has separately been reported to be building out a large Nvidia GPU footprint for its AI infrastructure, which makes the case for a parallel, internally controlled chip supply chain easier to understand.
On the Tesla side, the chip angle connects to friction the company has reportedly had with external chip partners. Tesla has been accused of pushing hard on pricing and terms with Samsung over its AI5 and AI6 chip orders, a dynamic that adds weight to the argument for Tesla wanting more in-house chip capacity rather than negotiating every generation of silicon with an outside foundry partner.
Wall Street’s reaction to the Terafab update has been measured rather than euphoric, reflecting the long timeline involved and the explicit uncertainty Musk attached to it, according to coverage tracking Tesla stock news. The dozens of analysts who track SpaceX closely tend to focus more on the company’s broader AI and infrastructure positioning than on any single hardware project, and Terafab is likely to be evaluated in that same broader context rather than as a standalone catalyst.
The Chip Shortage Backdrop Terafab Is Being Built Into
Terafab’s announcement does not happen in a vacuum. The broader AI hardware market has been dealing with persistent supply constraints across multiple chip categories throughout 2026, from GPUs to high-bandwidth memory. Micron has reported sharp increases in its own high-bandwidth memory sales this year, a sign of just how tight memory supply has become as AI infrastructure buildouts accelerate across the industry.
That shortage backdrop is likely part of why Musk is willing to accept the execution risk of building a first-time fab rather than simply continuing to compete for allocation from existing foundries. If chip supply remains constrained through 2027 and into 2028, having even a portion of internal production capacity, however modest relative to the terawatt-scale ambition, could meaningfully reduce Tesla and SpaceX’s exposure to future allocation squeezes.
Confirmed vs. Unconfirmed: What We Actually Know About Terafab
Given how much speculation tends to attach itself to any Musk-led manufacturing announcement, it helps to separate what has actually been confirmed from what remains unknown. The table below draws only on the details Musk shared on the October 20, 2026 call, as reported by Reuters and corroborated by Space.com and the South China Morning Post.
| Detail | Status |
|---|---|
| Volume production target date | Confirmed as a target: Q3 2028, possibly Q4 2028 |
| First test chip (SRAM) date and location | Confirmed as a target: roughly May/June 2027, Austin, Texas |
| Intended applications | Confirmed: Tesla Optimus and Cybercab chips; SpaceX space-based data center chips |
| Long-term output objective | Confirmed: approximately 1 terawatt per year of computing power |
| Manufacturing process node | Unconfirmed publicly |
| Wafer-start capacity / annual chip count | Unconfirmed publicly |
| Facility construction cost | Unconfirmed publicly |
| Production split between SpaceX and Tesla | Unconfirmed publicly |
Risks and Open Questions Around Terafab
Several open questions will determine whether Terafab becomes meaningful to the chip industry or stays an internal supply source for Tesla and SpaceX. First is simple execution: going from an SRAM test chip in 2027 to volume production of presumably more complex logic and memory chips in 2028 is a significant technical leap, one that established foundries with far more manufacturing experience have sometimes struggled to hit on schedule.
Second is financing and construction. Musk did not detail a construction cost or funding structure for Terafab in the remarks reported from the call, and building semiconductor fab capacity at any meaningful scale is capital-intensive even for well-funded companies. Third is talent: operating a leading-edge or near-leading-edge chip fab requires specialized engineering expertise that is in short supply across the entire industry, and SpaceX and Tesla would be competing for that talent against established foundries that are themselves expanding.
Finally, nothing in the current reporting suggests Terafab is intended to sell capacity to outside customers the way TSMC or Samsung Foundry do, which would keep its impact on the broader chip market limited even if it succeeds on its own terms.
Predictions: What Happens Next for Terafab
Based on the publicly confirmed details and the pattern of how Musk’s companies have historically handled aggressive manufacturing timelines, here are five reasonable expectations for how the Terafab story develops over the next two years.
- The May/June 2027 SRAM test chip milestone will likely become the next major public checkpoint investors and reporters track, since it is the nearest verifiable date on the current schedule.
- Expect the Q3 2028 volume-production date to shift at least once in subsequent earnings calls or investor updates, consistent with how Musk has handled comparable timelines for Starship and Full Self-Driving in the past.
- SpaceX and Tesla are likely to disclose additional specifics, such as process node details or capacity figures, only gradually, probably timed around major earnings calls or investor days rather than standalone announcements.
- Scrutiny of Terafab’s capital requirements will likely intensify as the project nears its 2027 test-chip milestone, especially given how much capital-intensive AI infrastructure spending is already underway across the industry.
- Analysts covering SpaceX and Tesla will probably continue treating Terafab as a secondary factor in their broader AI and infrastructure thesis for both companies, rather than a standalone catalyst, until it clears the 2027 test-chip milestone.
How Terafab Fits Into Musk’s Broader AI Strategy
Terafab is best understood as one piece of a larger pattern across Musk’s companies: an effort to control more of the AI and robotics supply chain rather than depend entirely on outside vendors. Tesla has pursued similar vertical integration in battery manufacturing, and SpaceX built out its own satellite and launch capacity rather than relying on third parties. A captive chip fab fits that playbook, even if the execution risk here is arguably higher, since chip manufacturing is among the most capital- and expertise-intensive industries in the world.
Whether Terafab succeeds on Musk’s stated timeline or slips, as several of his past manufacturing projects have, the attempt signals how central chip access has become to both companies’ growth plans. Tesla’s robotics ambitions and SpaceX’s AI buildout both depend on a steady supply of advanced silicon, and Terafab is Musk’s bet that building that supply in-house is worth the risk of a first-time fab operator chasing an aggressive schedule.
Frequently Asked Questions
What is Terafab?
Terafab is a chip manufacturing project shared between SpaceX and Tesla, aimed at producing chips for Tesla’s Optimus robots and Cybercab self-driving vehicles, as well as higher-power chips for SpaceX’s space-based data center plans.
When will Terafab begin volume chip production?
Elon Musk said on October 20, 2026 that Terafab is targeting the third quarter of 2028 for volume production, with the fourth quarter of 2028 also described as possible. Musk emphasized the date carries significant uncertainty.
What is the first milestone for Terafab?
The first major milestone is a functional SRAM test chip, targeted for approximately May or June 2027 at Terafab’s research facility in Austin, Texas.
Why do SpaceX and Tesla need their own chip factory?
Both companies have growing chip needs tied to AI and robotics: Tesla for its Optimus robots and Cybercab self-driving systems, and SpaceX for AI infrastructure and planned space-based data centers. Building in-house capacity would reduce their dependence on external foundries.
How much computing power is Terafab expected to produce?
Musk described a long-term production objective of approximately 1 terawatt per year of computing power dedicated to AI and robotics, though no date has been attached to that figure.
Is Terafab a confirmed, guaranteed project?
Musk explicitly said he puts “big error bars” on the 2028 timeline and acknowledged it is uncertain whether Terafab will hit that schedule. It should be treated as a stated target, not a confirmed outcome.
Will Terafab sell chips to companies outside of SpaceX and Tesla?
Nothing in the current public reporting indicates Terafab is intended to sell chip capacity to outside customers. It appears designed primarily to serve SpaceX and Tesla’s own internal needs.
What chip manufacturing process will Terafab use?
The specific manufacturing process node has not been publicly confirmed as of this reporting. Treat any specific process-node claims circulating elsewhere with caution until SpaceX or Tesla issue an official statement.
