Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Marvell completed its acquisition of Celestial AI on February 2, 2026. The deal brings Celestial’s Photonic Fabric optical-interconnect technology into Marvell’s Data Center Group, adding a scale-up connectivity bet to its broader AI infrastructure portfolio. The announced upfront consideration was about $3.25 billion, with additional share consideration tied to revenue milestones; the expected commercial payoff remains a management forecast, not reported revenue.
What happened, and when?
Marvell announced a definitive agreement to acquire Celestial AI on December 2, 2025. The FTC granted early termination of the applicable waiting period on January 21, 2026, and Marvell announced that the acquisition had closed on February 2, 2026.
Marvell placed Celestial in its Data Center Group. Its fiscal 2027 first-quarter filing included Celestial’s results from the February 2 closing date onward. The filing said the purchase-price allocation was preliminary, so some asset, liability, goodwill, and intangible-asset amounts could still change during the measurement period.
Sources: Marvell’s December 2, 2025 announcement, the FTC early-termination notice, and Marvell’s closing announcement.
#1 Best Overall
- ✅Powered by 26 Tera-Operations Per Second (TOPS) Hailo-8 AI Processor. 2.5W typical power consumption
- ✅Scalable, enabling simultaneous processing of multi-streams & multi-models
- ✅Enabling real-time, low latency and high-efficiency AI inferencing on the edge devices
- ✅Supports TensorFlow, TensorFlow Lite, ONNX, Keras, Pytorch frameworks
- ✅Supports Linux and Windows. Supports the temperature range of -40°C to 85°C
What does Celestial AI’s Photonic Fabric do?
Photonic Fabric is designed to use optical links to connect processors, accelerators, memory, and related components in large AI systems. Marvell describes it as a scale-up interconnect platform intended to operate at package, system, and rack levels. The goal is to move data between closely coordinated computing resources when electrical links face growing bandwidth, distance, and power constraints.
Marvell’s transaction materials said a single Photonic Fabric chiplet could deliver 16 Tbps of bandwidth. That is a company-stated capability in the transaction presentation, not independent evidence of production performance. Marvell also described possible future uses such as pooled-memory appliances and replacing some electrical die-to-die links in multi-die packages; these are potential applications, not proof that every use is already commercial.
Optical connectivity can offer potential bandwidth-density and energy-efficiency advantages, but those are design rationales rather than guarantees about total system cost or power in deployed systems. Products still need to meet practical requirements for reliability, yield, packaging, thermal management, manufacturing, and system integration.
Rank #2
- High-Performance AI Processing: The MX3 is designed to handle the most demanding AI computer vision workloads, delivering exceptional performance and efficiency.
- Flexible Integration: The MX3 can be easily integrated into your existing systems via its M.2 M-key form factor and support for Linux operating systems.
- Energy Efficient: The MX3 is designed to provide high performance while minimizing power consumption.
- Comprehensive Software Development Kit (SDK): The MX3 is supported by a comprehensive SDK that simplifies development and deployment.
- Hardware compatability: The MX3 is compatible with the PCI-SIG M.2 M-key 2280 Specification. It can be used with the Raspberry Pi 5 with a M-key 2280 HAT.
Source: Marvell’s transaction presentation.
How is scale-up different from scale-out?
| Term | What it connects | Relevance to this deal |
|---|---|---|
| Scale-out | Separate servers or systems across a broader network. | Marvell already has businesses spanning data-center connectivity, including scale-out optics and Ethernet switching. |
| Scale-up | Processors and memory within a large accelerated-computing system, potentially across a rack-scale architecture. | Celestial’s stated focus is optical connectivity for this increasingly demanding layer. |
As AI systems combine more accelerators, the links between them must carry more data with low latency and manageable power use. That does not make scale-up optics interchangeable with ordinary server-to-server transceivers: they address different connection problems. Marvell’s wider portfolio also includes custom silicon, electro-optics, switches, and scale-out products; Celestial does not supply all those functions by itself.
Source: Marvell’s fiscal 2027 first-quarter earnings release.
Why did Marvell want Celestial AI?
- Address a scaling bottleneck: As more accelerators work together, moving data among them becomes a larger part of system design. Marvell is betting optical links can help address bandwidth and reach demands that become difficult for electrical interconnects at very large scale.
- Broaden its AI infrastructure portfolio: Celestial adds a scale-up optical layer alongside Marvell’s custom silicon, switching, electro-optics, and other data-center connectivity businesses.
- Own the platform: Bringing Celestial’s technology and team in-house gives Marvell control of a potentially important architecture, while also putting responsibility for integration and execution on Marvell.
- Build on reported market interest: Marvell said Celestial was engaged with multiple hyperscalers and ecosystem partners. The acquisition announcement did not name those organizations, and engagement alone does not establish a binding purchase contract.
The strategic upside depends on translating an optical architecture into reliable, manufacturable products and integrating them into customers’ systems. The transaction is therefore a bet on both technology and commercialization, not simply on the general growth of AI computing.
Sources: Marvell’s transaction announcement and closing announcement.
Recommended Free Tools
What were the deal terms?
| Stage or component | Reported terms |
|---|---|
| Announced upfront consideration | About $3.25 billion at signing: approximately $1.0 billion cash and 27.2 million Marvell shares valued at about $2.25 billion using the specified 10-trading-day volume-weighted average price. |
| Potential contingent consideration | Up to approximately 27.2 million additional Marvell shares, valued at up to about $2.25 billion under the announcement’s reference valuation, subject to revenue milestones. Including this maximum contingent amount, the potential total was roughly $5.5 billion; it was not all paid at closing. |
| Earnout milestones | The first milestone, one-third of the contingent consideration, required at least $500 million of cumulative revenue by the end of Marvell fiscal 2029. The full earnout required cumulative revenue above $2.0 billion by the end of fiscal 2029. |
| Disclosed at closing | Marvell’s fiscal 2026 annual report said it paid about $1.3 billion gross cash, or about $1.0 billion net of approximately $300 million of cash acquired, and issued about 24.5 million shares. Additional cash and shares could still be owed if the specified milestones are achieved. |
The signing and closing figures describe different things. The $3.25 billion announced upfront value used an estimated share value based on a reference VWAP; the later annual-report figures reflect cash and shares actually delivered at closing. The difference is not evidence that the terms were simply paid in the original estimated amounts. Contingent consideration is separate and depends on future revenue performance.
Sources: December 2025 Form 8-K and Marvell fiscal 2026 annual report.
Rank #4
- ✅Powered by 26 Tera-Operations Per Second (TOPS) Hailo-8 AI Processor. 2.5W typical power consumption
- ✅Scalable, enabling simultaneous processing of multi-streams & multi-models
- ✅Enabling real-time, low latency and high-efficiency AI inferencing on the edge devices
- ✅Supports TensorFlow, TensorFlow Lite, ONNX, Keras, Pytorch frameworks
- ✅Supports Linux and Windows. Supports the temperature range of -40°C to 85°C
What revenue does Marvell expect, and when?
Marvell’s announced timetable put initial revenue contributions in the second half of fiscal 2028, with a $500 million annualized run rate in the fourth quarter of fiscal 2028 and a $1 billion annualized run rate in the fourth quarter of fiscal 2029. The transaction presentation also forecast non-GAAP earnings accretion in the second half of fiscal 2028. These are management projections, not reported results or guaranteed customer commitments.
“Annualized run rate” is not the same as revenue earned during a quarter or a full fiscal year. It extrapolates a quarterly pace to an annual rate; a $500 million annualized rate in a quarter does not mean Celestial necessarily generated $500 million of revenue in that quarter.
Marvell fiscal years end around the Saturday nearest January 31, so fiscal 2028 and fiscal 2029 should not be read as calendar 2028 and calendar 2029. The expected lag between the February 2026 closing and meaningful revenue is material to assessing execution timing.
Best Value
- DEEPX DX-M1M NPU: Powered by the DEEPX DX-M1M neural processing unit, purpose-built for efficient on-device AI inference workloads.
- COMPACT M.2 2242 FORM FACTOR: Fits the standard M.2 2242 slot, making it easy to integrate into embedded systems, edge devices, and compact computing platforms.
- EDGE AI ACCELERATION: Designed to accelerate deep learning inference at the edge, enabling real-time AI applications without relying on cloud connectivity.
- RADXA AICORE MODULE: The Radxa AICore DX-M1M delivers a plug-and-play AI compute solution ideal for robotics, smart cameras, and industrial automation.
- WARRANTY AND ORIGIN: Backed by a 1-year manufacturer warranty and crafted with quality components for reliable long-term performance in demanding environments.
Sources: Marvell’s transaction announcement and transaction presentation.
What has changed financially since closing?
Marvell said the transaction reduced its cash balance by approximately $1 billion, lowering expected future interest income by approximately $38 million annually. That interest-income effect is distinct from the purchase consideration and from any future revenue Celestial may generate.
Marvell’s fiscal 2027 first-quarter report showed company-wide revenue of $2.418 billion, up 28% year over year. It cited demand in several AI-related areas, including scale-up optical solutions for NPO and CPO applications. Those results cover Marvell’s wider business; they do not establish that Celestial alone generated the company’s growth or quantify Celestial’s contribution.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Sources: Marvell’s closing announcement and fiscal 2027 first-quarter earnings release.
What are the main risks to the acquisition thesis?
- Technology-to-production execution: Optical designs must work at scale across reliability, yield, packaging, thermal, manufacturing, and system-integration requirements.
- Commercialization timing: Marvell’s forecast places initial revenue contributions in the second half of fiscal 2028, leaving a substantial period after closing before the expected contribution begins.
- Customer visibility: Marvell cited hyperscaler and ecosystem engagement but did not identify the customers in its acquisition announcement. Public information there does not establish the scale, timing, or terms of customer commitments.
- Potential dilution: If the revenue milestones are met, Marvell may issue substantial additional shares. The earnout can reward commercial success but would dilute existing shareholders.
- Integration and retention: Marvell identified retaining employees, integrating the business, maintaining customer relationships, and implementing post-acquisition plans as risks.
- Competition and forecast risk: Scale-up connectivity is contested across networking, switching, optical, silicon-photonics, co-packaged-optics, and custom-silicon approaches. Marvell’s run-rate targets are forecasts, not contracted sales.
Sources: transaction announcement, closing announcement, and Marvell’s fiscal 2027 first-quarter filing.
What should readers take from the deal?
Marvell has completed a strategic acquisition aimed at optical scale-up connectivity for large AI systems. Celestial gives Marvell a technology platform and team in an area the company sees as important, but the financial case depends on manufacturing, integration, customer adoption, and the delivery of management’s later fiscal-year targets. The transaction is best understood as a long-term infrastructure bet whose potential upside comes with execution risk and possible share dilution.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
