SpaceX Super Intelligence (SpaceXSI) Memphis supercomputing campus is approaching an extraordinary scale. Still, the gap between reported GPU deployments and available electrical capacity raises important questions about how much of that computing power can actually come online.

The race for artificial intelligence infrastructure is increasingly defined by metrics involving hundreds of thousands of GPUs, gigawatts of electricity, and multibillion-dollar capital investments. However, beneath these substantial deployment announcements lies a more critical engineering challenge: determining the proportion of hardware that can be powered and operated concurrently.

The SpaceXSI Colossus 2 campus in Memphis, Tennessee, serves as a significant case study. As of November 3, 2026, data derived from infrastructure updates indicates five distinct installations at the campus, comprising approximately 1.19 million Blackwell GPUs, including both NVIDIA GB200 and GB300 accelerators. While this represents an unprecedented hardware scale, Epoch AI’s research estimates the facility's current IT power capacity at approximately 946 megawatts, with projections for growth to 1,531 MW by 2027.

These figures warrant closer scrutiny. An analysis of an October 31 snapshot, which identified 860,000 GB300 GPUs, suggests a nominal IT power requirement of approximately 1.64 gigawatts, exceeding Epoch AI’s estimated capacity by roughly 700 MW. Although the updated November 3 figures shift the scale of the deployment, the fundamental discrepancy remains. The divergence between reported GPU inventory and the capacity for energized, operational hardware has emerged as a definitive challenge in the development of hyperscale AI infrastructure.

Five installations, nearly 1.2 million Blackwell GPUs

Today’s breakdown identifies five named installations within the Colossus 2 development.

The reported configuration is:

• Macrohard: Approximately 330,000 accelerators, including 110,000 GB200s and 220,000 GB300s.
• Macroharder: Approximately 220,000 GB300s.
• Minihard: Approximately 220,000 GB300s.
• Gun Range: Approximately 220,000 GB300s, targeted for operational status by the end of November.
• Spam: Approximately 200,000 GB300s, intended primarily for inference workloads in December.

Combined, those installations represent approximately 1.19 million Blackwell accelerators, including approximately 1.08 million GB300s and 110,000 GB200s.

The distinction between these installations matters.

Musk's October 31 update described Macroharder and Minihard as training-oriented installations using 800-gigabit networking. Gun Range is expected to use a similar configuration.

Spam, meanwhile, is reportedly intended for inference workloads, with a simpler networking configuration and substantially less fiber.

That distinction offers an interesting glimpse into how the campus is being organized around different computing workloads.

Large-scale model training requires extensive communication between accelerators. Distributed training systems depend on high-bandwidth, low-latency networking, efficient collective communication, and carefully balanced compute and memory resources.

Inference has different performance requirements. Depending on model architecture, batching strategy, and serving configuration, inference clusters can prioritize throughput, latency, memory capacity, and utilization rather than precisely reproducing the networking architecture of a massive training installation.

The decision to configure different buildings for different workloads is therefore technically significant.

But there’s an important qualification: These figures are based on Musk's reported updates and secondary coverage, not an independently audited inventory of operational GPUs.

The distinction between announced capacity, delivered hardware, installed systems, and energized computing resources remains unresolved.

The 700 MW discrepancy deserves attention

The earlier 860,000-GB300 figure provides a useful engineering benchmark.

An NVIDIA GB300 NVL72 rack contains 72 Blackwell Ultra GPUs and 36 Grace CPUs. OEM-related figures cited in the research brief place nominal rack power requirements at approximately 132 to 142 kilowatts, with peak requirements approaching 155 kilowatts.

Using approximately 137 kilowatts per rack produces the following calculation:

860,000 GPUs divided by 72 GPUs per rack produces approximately 11,944 racks.

At 137 kilowatts per rack, the resulting nominal IT power requirement approaches 1.64 gigawatts.

Epoch AI estimates Colossus 2's current IT power capacity at approximately 946 MW.

The difference is approximately 700 MW.

This is a derived engineering estimate, not a measurement of Colossus 2's actual electrical consumption. It assumes that the reported GPU count can be translated into fully populated GB300 NVL72 racks operating at nominal rack power.

It also does not mean all 860,000 GPUs were simultaneously operational when the estimate was made.

The newer 1.19 million figure should not simply be substituted into the same calculation. It includes GB200 accelerators, different deployment stages, and installations that were still awaiting operational status.

Nevertheless, the original calculation identifies a significant infrastructure question.

Where does the difference between reported hardware and available electrical capacity come from?

Epoch AI's numbers require careful interpretation

Epoch AI's infrastructure database is valuable because it attempts to measure data-center capacity using a combination of company disclosures, satellite imagery, cooling infrastructure, and power-generation estimates.

But its figures are estimates, not independently verified electrical meter readings.

The organization's approximately 946 MW Colossus 2 estimate represents IT power capacity. It should not be confused with actual electricity consumption at any particular moment or total facility power, which also includes cooling and other infrastructure overhead.

Epoch states that approximately 80 percent of its IT power estimates fall within a factor of 1.4 of the true value.

Applied to 946 MW, that produces an approximate upper boundary of 1.32 GW.

Even that figure remains below the approximately 1.64 GW derived from the earlier 860,000-GB300 snapshot.

The discrepancy cannot comfortably be dismissed as ordinary estimation uncertainty alone.

Deployment timing is a more plausible explanation.

Epoch's estimate reflects an earlier infrastructure assessment, while Musk's reported hardware figures include installations at different stages of deployment.

That distinction is critical. GPUs can be delivered and installed before the electrical distribution, cooling, and supporting infrastructure required for full operation are ready.

The numbers can therefore describe different realities without necessarily contradicting one another.

Still, Epoch AI should clarify precisely which estimation methodology it used for Colossus 2 and how its model accounts for the campus's unusual generation arrangements.

Electricity is becoming the limiting factor

Colossus 2's power infrastructure provides perhaps the most important context for understanding the deployment.

The campus has relied on temporary natural-gas turbines to accelerate its power availability.

A permanent 1.2-gigawatt power plant in Southaven, Mississippi, is expected around mid-2027, according to sources.

Meanwhile, 69 temporary turbines are subject to an agreed removal order extending through July 2027.

Reported battery storage capacity involves approximately 720 Tesla Megapack containers, representing an estimated 2.8 to 3.3 gigawatt-hours.

Battery storage can help balance demand, provide backup capacity, and manage short-term fluctuations. However, batteries are not an independent source of sustained electricity. They must be charged, and their usefulness depends on available generation and the duration and rate of discharge.

For an AI supercomputing facility operating thousands of high-density racks, dependable electrical generation is essential.

A large GPU installation cannot be evaluated simply by counting its accelerators. Electrical distribution equipment, transformers, substations, backup systems, cooling infrastructure and grid interconnections all influence how much computing capacity can be delivered.

The planned Southaven plant therefore represents a major part of the campus's long-term expansion strategy.

Epoch AI projects Colossus 2's IT power capacity will increase from approximately 946 MW to approximately 1,531 MW in 2027.

That would represent a substantial improvement in the facility's ability to support additional computing hardware.

But projected capacity is not operational capacity.

Until the generation and supporting infrastructure are commissioned, the difference between planned computing resources and available computing resources remains an important consideration.

Training and inference create different infrastructure demands

Another interesting development is the reported separation between training and inference installations.

Macroharder, Minihard, and the planned Gun Range configuration reportedly use 800-gigabit networking for training workloads.

That is a significant networking requirement.

Large distributed AI training systems must continuously exchange information between accelerators. Depending on the parallelization strategy, communication can involve all-reduce operations, gradient synchronization, parameter exchange, and other collective operations.

As accelerator counts increase, network topology, congestion control, optical connectivity, and communication efficiency become increasingly important.

Adding GPUs without providing sufficient network bandwidth can produce diminishing returns. Accelerators may spend valuable time waiting for data or synchronization rather than performing useful computation.

Inference introduces another set of challenges.

The Spam installation is reportedly configured for inference and is expected to require substantially less fiber than the training-oriented buildings.

That makes sense as an architectural distinction, although actual networking requirements will depend on model size, serving topology, batching, latency targets, and the number of concurrent users.

The broader point is that a 1.19 million GPU campus is not necessarily one uniform supercomputer.

It is better understood as a collection of large computing installations with different workload requirements, networking configurations and operational timelines.

That distinction matters when comparing theoretical aggregate computing capacity with the performance available to actual users.

Three different definitions of the world's largest AI data center

Epoch AI's rankings also illustrate how difficult it has become to establish a single winner in the AI infrastructure competition.

Depending on the measurement, different facilities lead.

Colossus 2 leads in reported GPU inventory and Epoch AI's estimated compute capacity, which the organization places at approximately 1.4 million H100-equivalents.

New Carlisle, an Anthropic-Amazon infrastructure site, leads in current IT power capacity at approximately 1,037 MW.

Microsoft's Fairwater facility in Wisconsin leads Epoch AI's planned IT power ranking at approximately 2,263 MW, with a projected completion milestone in the second quarter of 2028.

These measurements cannot be treated as interchangeable.

GPU count measures accelerator inventory. H100-equivalents attempt to normalize compute capacity across different accelerator generations. IT power capacity measures the electrical resources available to computing equipment.

A facility can lead in one category without leading in the others.

The industry needs more consistent disclosure around what these enormous numbers actually represent.

How many accelerators have been delivered? How many have been installed? How many are energized? How much electrical capacity is available? And how much of that capacity is being used for productive computing?

Without those distinctions, comparisons between competing AI infrastructure projects risk becoming marketing exercises rather than meaningful technical assessments.

The next milestones will reveal whether the numbers hold up

Several upcoming milestones will help determine whether Colossus 2 can convert its reported hardware inventory into operational computing capacity.

The first is Gun Range, which Musk says is tracking toward operational status by the end of November.

The second is Spam, which is expected to enter operation in December.

The third is the approximately 1.2 million Blackwell target for the end of 2026.

The fourth is Epoch AI's next infrastructure update.

If its Colossus 2 estimate moves toward the projected 1.5 GW level, that would provide evidence that additional power and cooling infrastructure is becoming available.

The planned Southaven power plant, expected around mid-2027, will be another important milestone.

But several questions remain unanswered:

• How many GPUs are physically installed versus delivered but awaiting deployment?
• How many accelerators are energized and available for sustained workloads?
• What is the actual electrical demand of the operational computing fleet?
• How much capacity is constrained by generation, transmission, substations, or cooling?
• Can the networking infrastructure sustain the performance expected from such a large distributed training environment?

These are questions that SpaceXSI and its infrastructure partners should be prepared to address.

The real competition is no longer just about GPUs

The Colossus 2 project highlights the rapid evolution in how we define supercomputing scale. The industry has progressed beyond merely tracking accelerator counts and theoretical floating-point performance; modern AI infrastructure now relies on the complex integration of compute density, memory bandwidth, high-speed interconnects, cooling systems, power distribution, and reliable generation. 

Any accelerator lacking sufficient power, a rack hindered by inadequate cooling, or a cluster restricted by insufficient networking fails to deliver its full computational potential. These are not secondary engineering details, but rather fundamental requirements of the entire system. While Colossus 2 may eventually reach the scale its developers envision, the industry’s current focus on hardware procurement must be balanced by an appreciation for the underlying engineering realities required for operation. Ultimately, the next phase of the AI infrastructure race may not be determined by the volume of hardware acquired, but by the ability to provide the necessary power, cooling, and networking to sustain it at scale. In this emerging era of gigawatt-level supercomputing, the most critical metric may shift from total GPU inventory to the capacity for operational, energized hardware.

The proposed acquisition of Lumentum brings custom silicon, high-speed networking, lasers, and photonics under one roof. But as AI clusters push copper connectivity toward its physical limits, Marvell is making an enormous bet that optical interconnects will define the next generation of computing infrastructure.

The artificial intelligence industry has invested billions in increasingly powerful processors, with NVIDIA GPUs, custom accelerators, high-bandwidth memory, and massive computing clusters scaling modern data center workloads. However, as these systems expand, the challenge of facilitating efficient data movement between processors has become increasingly acute. 

Marvell Technology posits that the solution lies in optical interconnects. On November 2, the company announced a definitive agreement to acquire the optical technology manufacturer Lumentum in an all-stock transaction valued at approximately $115 billion, inclusive of debt. This deal ranks among the largest technology acquisitions in history and signifies a concerted effort to integrate electrical connectivity, custom silicon, and advanced photonics under a unified corporate architecture. 

The strategic rationale is substantial; as AI infrastructure evolves beyond individual racks toward highly interconnected computing systems, conventional electrical connections face significant limitations regarding bandwidth, signal integrity, power consumption, and latency. By acquiring Lumentum, Marvell seeks to secure critical technologies to address these bottlenecks. Nevertheless, this acquisition prompts a pivotal question: is Marvell successfully positioning itself at the forefront of a fundamental architectural shift, or is it paying a premium for technology that may be approaching a cyclical peak in demand?

The networking problem behind the AI supercomputer boom

Modern AI supercomputers are not simply collections of powerful GPUs. They are highly integrated computing environments in which thousands of processors must exchange data continuously.

Training large models requires frequent communication between accelerators. Distributed workloads depend on moving parameters, gradients, activations, and intermediate results across high-speed interconnects. Inference infrastructure introduces its own demands as enormous numbers of requests compete for computing resources and memory bandwidth.

The performance of an individual accelerator is therefore only part of the equation.

A system can contain extraordinarily powerful processors and still fail to deliver proportional performance improvements if communication between those processors becomes the limiting factor.

This is where networking architecture becomes critical.

Copper-based electrical connections remain essential throughout data center infrastructure, but increasing transmission rates and shorter electrical signal budgets create difficult engineering challenges. Higher signaling speeds require careful management of signal loss, equalization, power consumption, and thermal characteristics.

As AI clusters scale, these limitations become increasingly consequential.

Optical technologies offer a different approach. Instead of transmitting electrical signals over copper conductors, optical interconnects convert electrical information into light and transmit it through optical media.

The potential advantages include higher bandwidth over appropriate distances, improved signal integrity, and opportunities to reduce the energy required for data movement.

However, optical connectivity is not a universal replacement for copper. The economics, conversion overhead, packaging complexity, reliability, and deployment requirements still matter.

The industry is increasingly interested in combining electrical and optical technologies at progressively closer points within the computing system.

That transition is the foundation of Marvell's acquisition strategy.

Why combining Marvell and Lumentum matters

The two companies occupy complementary positions in the data center connectivity supply chain.

Marvell brings networking processors, digital signal processors (DSPs), switching technology, and custom application-specific integrated circuits for cloud computing customers.

Lumentum brings the photonic components that make optical communication possible, including high-performance lasers, optical transceivers, optical circuit switching technologies, and compound-semiconductor manufacturing capabilities.

Its indium phosphide (InP) technology is particularly important because InP is a key material used in high-performance optical devices.

The combination could allow Marvell to coordinate more closely the electrical and optical components responsible for moving data through increasingly complex AI systems.

That matters because designing an optical link is not simply a matter of selecting a faster laser.

The electrical interface, DSP, optical engine, laser characteristics, packaging, thermal behavior, and system-level integration all influence the performance of the final solution.

Bringing these technologies together could provide greater control over the complete connectivity architecture.

Marvell has already been moving in this direction. The company completed its $3.5 billion acquisition of Celestial AI in February, adding another important photonics technology portfolio to its expanding AI infrastructure strategy.

Lumentum would substantially increase that commitment.

The opportunity is to move beyond selling individual components and toward providing more completely integrated, optimized, and validated connectivity solutions.

For hyperscale data center operators, that could eventually simplify the development and deployment of increasingly complex AI infrastructure.

Co-packaged optics could change the architecture

One of the most important technologies behind this acquisition is co-packaged optics (CPO).

Traditional optical networking generally relies on pluggable transceivers positioned at the front of networking equipment. Electrical signals travel from processors or switching ASICs toward optical modules, where they are converted into light for transmission.

This architecture has served data centers well, but increasing bandwidth requirements are putting pressure on the electrical connections between switching silicon and optical modules.

CPO changes that arrangement.

Rather than keeping optical engines in separate pluggable modules, co-packaged optics places optical components much closer to the switching or processing silicon within a shared package.

Near-packaged optics (NPO) represents another approach, positioning optical engines close to the processor or switch without necessarily integrating them into the same package.

Both approaches seek to shorten the electrical connections required to reach optical conversion.

The potential benefits include reduced electrical transmission distances, improved bandwidth density, and opportunities to lower connectivity power consumption.

For large AI computing installations, these improvements could become increasingly important.

Consider the architecture of a large accelerator cluster. Thousands of GPUs or custom accelerators must communicate through a hierarchy of high-speed links and switches. As the number of endpoints increases, the networking fabric must deliver more aggregate bandwidth while managing latency, power, and physical complexity.

Optical integration closer to the switching silicon could help address some of those constraints.

But CPO is not a magic solution.

Manufacturing yields, laser reliability, thermal management, fiber attachment, serviceability, packaging costs, and field replacement procedures remain important engineering considerations.

A failed pluggable transceiver can generally be replaced without replacing an entire switching system. More tightly integrated optical architectures introduce different maintenance and reliability trade-offs.

The question is not whether CPO is technically promising. It is whether manufacturers can deliver reliable, economically competitive systems at the production volumes demanded by hyperscale customers.

Marvell is betting that it can.

NVIDIA: Strategic partner, customer, investor, or competitor?

Perhaps the most interesting aspect of this transaction is NVIDIA's position.

Earlier in 2026, NVIDIA invested $2 billion in each company.

In Lumentum's case, the investment was accompanied by a multibillion-dollar purchase commitment for laser components and support for a new U.S. manufacturing facility.

That makes NVIDIA more than an outside observer of the transaction.

It has financial interests in both companies, relies on an expanding ecosystem of networking and optical technologies, and is developing its own photonics capabilities.

NVIDIA's networking strategy increasingly extends beyond the GPU itself. Its AI infrastructure platforms depend on high-performance networking, switching, and interconnect technologies to connect accelerators into large computing systems.

The company's own co-packaged optics developments could eventually place it in direct competition with suppliers such as the combined Marvell-Lumentum business.

This creates an interesting strategic relationship.

NVIDIA could benefit from a stronger Marvell capable of supplying more advanced connectivity technologies to the broader AI infrastructure market.

At the same time, NVIDIA has every incentive to maintain control over its own networking architecture and supply chain.

The question is whether Marvell becomes an increasingly important supplier within NVIDIA's ecosystem or develops into a competitor in some of the same optical networking markets.

That distinction could become important as optical technologies move closer to the processor package.

The Broadcom challenge

The acquisition also strengthens Marvell's competitive position against Broadcom.

Broadcom has established a substantial presence in custom AI silicon and data center networking. Its relationship with Google has made it an important participant in the development of custom accelerators designed for hyperscale computing environments.

Marvell has been pursuing the same opportunity.

Its August agreement with Google, involving custom silicon with potential sales of up to $120 billion through fiscal 2033 and a right for Google to purchase a $12.2 billion stake, demonstrated the scale of its ambitions.

Adding Lumentum's optical technology could give Marvell a broader portfolio spanning custom processors, electrical connectivity, and optical components.

That is strategically attractive.

Hyperscale customers increasingly want suppliers capable of helping them optimize entire systems rather than simply delivering isolated chips.

However, owning more components does not automatically guarantee a better product.

The combined company will have to demonstrate that its integrated solutions deliver measurable advantages in bandwidth, power efficiency, manufacturing economics, and deployment flexibility.

Broadcom is unlikely to surrender its position without a fight.

The financial risk behind the optical opportunity

The strategic argument is compelling. The financial commitment is considerably harder to ignore.

Marvell is acquiring Lumentum through a fixed exchange ratio of 4.5 Marvell shares for every Lumentum share.

The announced $1,132 per-share figure represents a 23.9% premium based on October 28 closing prices, before market speculation about the transaction.

But that figure is not a guaranteed cash-equivalent purchase price.

Because the transaction is entirely stock-based, the actual value received by Lumentum shareholders will fluctuate with Marvell's share price until closing.

The distinction became particularly important following the announcement.

Monday's Marvell close is reported as $171.41 by TipRanks, marking a severe 12.72% drop on massive trading volume (implying ~$771 per Lumentum share), not $183.25 as Barron's had it (~$825). The premium-inversion conclusion holds under either figure.

In other words, the headline premium had effectively inverted.

That does not necessarily invalidate the acquisition. Lumentum shareholders will own approximately 33% of the combined company and participate in its future performance.

But it illustrates the risks associated with using a volatile stock as acquisition currency.

The companies also project a combined addressable market of approximately $550 billion, more than $1 billion in annual run-rate cost savings, and first-year post-closing adjusted earnings-per-share accretion.

Those are company projections, not independently established outcomes.

The acquisition will have to deliver considerably more than an impressive presentation to justify its extraordinary valuation.

The optical supply chain could become a strategic battleground

There is another dimension worth watching: manufacturing capacity.

Advanced optical components are not interchangeable commodities that can simply be produced by any semiconductor factory.

Specialized materials, photonic device fabrication, laser manufacturing, packaging, and precision assembly all contribute to the production process.

Lumentum's indium phosphide manufacturing capabilities could therefore represent an important strategic asset.

The company has reported exceptionally strong demand for its products, including significant commitments associated with near-packaged and co-packaged optics.

Its latest reported quarterly revenue reached approximately $1.26 billion, with management forecasting substantially higher revenue in the following quarter.

The company has also indicated that demand for certain AI data center optical products extends into early 2029.

These figures suggest that optical supply is becoming a critical consideration in AI infrastructure expansion.

However, supply shortages can create unusually attractive market conditions that do not necessarily last.

New manufacturing capacity, competing technologies, changes in customer architectures, and adjustments in hyperscaler capital spending could alter the economics of optical components.

Marvell is committing enormous resources to a market whose long-term growth prospects may be substantial but whose near-term pricing and supply conditions could change.

There is also a regulatory dimension. Reuters reported that Lumentum could benefit from a proposed U.S. restriction on new Chinese optical transceivers. The proposal's specific regulatory mechanism, scope, and implementation timeline remain unclear. It should not be confused with an enacted policy.

A long road to integration

The transaction is expected to close in the first half of 2028, subject to shareholder approvals and regulatory clearances.

That leaves considerable time for market conditions, technology priorities, and competitive relationships to change.

Until closing, the companies will continue operating independently.

The combined organization will eventually have to integrate product roadmaps, engineering teams, manufacturing operations, customer relationships, and supply chains.

One particularly important question concerns Lumentum's existing customers.

The company serves a broad optical networking market, and some customers may compete directly with Marvell in custom silicon and networking.

Maintaining customer confidence while integrating the businesses will be essential.

The acquisition's success will depend not simply on combining technologies but on convincing customers that the resulting products will remain competitive, reliable, and available.

The real test: Does light make AI computing better?

Marvell's proposed acquisition of Lumentum represents a significant milestone in the evolution of high-performance computing. For years, the industry has primarily prioritized advancements in processor performance, accelerator density, memory capacity, and overall computing throughput. While these investments remain critical, the next generation of AI infrastructure will increasingly depend on the efficiency of data movement between processors, memory systems, racks, and data centers. 

Optical connectivity is poised to become a defining component of this architecture, and the integration of Marvell’s electrical connectivity expertise with Lumentum’s photonics capabilities offers a unique opportunity to accelerate this transition. However, a $115 billion acquisition does not inherently guarantee the projected economic returns. The true measure of success will be demonstrated through tangible improvements in bandwidth-per-watt efficiency, reduced system-level communication costs, reliable high-volume manufacturing, and enhanced performance across real-world AI workloads. 

The industry requires more than compelling technical demonstrations and ambitious market projections; it necessitates commercially viable systems that render large-scale computing faster, more efficient, and more cost-effective. Marvell is making one of the most substantial bets to date that the future of AI infrastructure will be defined not only by the speed of processors but by the light connecting them. For the supercomputing trade, this underscores a fundamental shift: the next performance revolution may depend as much on the efficacy of data transmission as it does on raw computational power.