Revision #1

On April 13, 2026, Credo Technology Group announced a definitive agreement to acquire DustPhotonics, an Israeli developer of silicon photonics integrated circuits, for $750 million in cash plus approximately 0.92 million shares of Credo common stock [1]. Including contingent consideration of up to 3.21 million additional shares tied to financial milestones, the deal could be worth as much as $1.3 billion [2]. The transaction is expected to close in the second quarter of 2026, pending regulatory approvals [1].

The market's initial verdict was favorable: Credo shares jumped roughly 10-12% in after-hours trading, reaching approximately $143 per share [3][4]. But beneath the headline numbers lies a more complex story about the economics of AI infrastructure, the strategic imperatives driving semiconductor consolidation, and whether $750 million for a 70-person startup with undisclosed revenue constitutes a bargain or a premium.

The Deal in Financial Context

Credo enters this acquisition from a position of unusual strength. The company reported $407 million in revenue for Q3 of fiscal year 2026 (ended January 2026), representing 201.5% year-over-year growth and 51.9% sequential growth [5]. Its trailing twelve-month revenue stands at approximately $1.06 billion [6]. Credo ended the quarter with $1.3 billion in cash and short-term investments [5], meaning the $750 million upfront cash payment represents roughly 58% of its liquid reserves.

Source: Credo Technology investor relations

Data as of Apr 14, 2026CSV

At a market capitalization of approximately $24.8 billion as of mid-April 2026 [7], the $750 million base price amounts to about 3% of Credo's equity value — a manageable figure for an acquisition of this strategic importance. The company's GAAP gross margin of 68.5% [5] provides the operating leverage needed to absorb DustPhotonics' operations without immediate margin compression.

But relative to DustPhotonics' own financials, the price demands scrutiny. The startup raised approximately $163 million in total venture funding from investors including Greenfield Partners, Sienna Venture Capital, Exor Ventures, Atreides Management, Key1 Capital, and Discount Capital [8][9]. No public revenue figures exist for DustPhotonics. Credo has stated that the combined portfolio of ZeroFlap Optical Transceivers, Optical DSPs, and Silicon Photonics products is expected to generate more than $500 million in optical revenue in fiscal 2027 [1]. If DustPhotonics contributes a meaningful fraction of that target — say $100 million to $150 million — the implied forward revenue multiple of 5x to 7.5x would be within the range of recent optical semiconductor transactions, though on the higher end given the company's startup-stage profile.

How $750 Million Compares to the Optical M&A Landscape

The Credo-DustPhotonics deal is modest by the standards of optical networking consolidation. Cisco initially bid $2.6 billion for coherent optics specialist Acacia Communications in 2019 before ultimately paying $4.5 billion to close the deal in 2021 [10][11]. The merger of II-VI Incorporated and Coherent Corp in 2022 was valued at approximately $7 billion [12]. Most recently, NVIDIA committed $4 billion in strategic investments split between Lumentum and Coherent in early 2026, a move CEO Jensen Huang described as securing photonics as "a core pillar of AI infrastructure" [13][14].

Source: Company filings and press releases

Data as of Apr 14, 2026CSV

In that context, $750 million for a pre-revenue or early-revenue silicon photonics company with a specific niche — photonic integrated circuits (PICs) for optical transceivers — is a comparatively contained bet. The contingent milestone structure further limits Credo's downside risk: the additional 3.21 million shares only vest if DustPhotonics hits specific financial targets [1].

What DustPhotonics Brings to the Table

DustPhotonics was founded in 2017 in Modi'in, Israel, by a team with deep optical and semiconductor experience: Ben Rubovitch (formerly of Mellanox), Dr. Kobi Hasharoni and Amir Geron (both from Compass Networks), and Yoel Chetrit (formerly Intel) [15]. The company's chairman is Avigdor Willenz, one of Israel's most prominent chip industry figures, who previously sold Galileo Technologies to Marvell for $2.7 billion in 2001 and Annapurna Labs to Amazon for $370 million in 2015 [16][17].

Current CEO Ronnen Lovinger, appointed in 2021 following a strategic pivot toward chip development, has led the company through its transition from optical module design to silicon photonics PIC production [15].

The company operates a fabless model with approximately 70 employees [18]. Its product line includes the Carmel series of silicon photonics PICs supporting data rates of 400 Gbps, 800 Gbps, and 1.6 Tbps [19]. Its most advanced product, the Tamar200, is described as the industry's first merchant 1.6T-2xFR4 PIC, supporting two transmit channels of 800G-FR4 with each optical lane operating at 200 Gbps [19].

This technology fills a specific gap in Credo's portfolio. Credo has built its business on Active Electrical Cables (AECs) and SerDes (serializer/deserializer) IP — the electronic components that convert parallel data into serial streams for high-speed transmission over copper cables [1]. But as AI cluster interconnect distances grow beyond the 5-7 meter reach of copper, optical connectivity becomes essential. Rather than spending years developing silicon photonics capabilities internally, Credo chose to acquire a team and technology platform already shipping silicon to customers.

CEO William Brennan described the acquisition as "a defining step in Credo's strategy to lead across the full spectrum of AI connectivity" [4], positioning the combined company with a vertically integrated stack spanning SerDes, Digital Signal Processing, Silicon Photonics, and system integration for both scale-out and scale-up AI networks [1].

The Customer Question

DustPhotonics has secured design wins with hyperscale cloud customers, though the company has not publicly disclosed a full client list [19]. Israeli business publication Globes identified Amazon as a customer of both DustPhotonics and Credo [2]. Reports from Calcalist Tech indicate that Intel, NVIDIA, and Amazon were among the parties that explored acquiring DustPhotonics [15], suggesting the startup's technology has been validated at the highest levels of the AI infrastructure supply chain.

The presence of multiple potential acquirers implies the $750 million price was set through competitive dynamics rather than a distressed sale. While the broader venture capital market for photonics startups has tightened since 2022, DustPhotonics' position in the AI infrastructure supply chain — directly serving hyperscaler demand — insulated it from the worst of the funding drought. The company's last known funding round was a $24 million Series B follow-on in February 2024 [9].

The AI Dependency Risk

The business case for this acquisition rests heavily on continued growth in AI cluster interconnect demand. Credo's own revenue trajectory — from $58.8 million in Q1 FY2025 to $407 million in Q3 FY2026 [5] — is driven almost entirely by hyperscaler spending on AI infrastructure. The optical interconnect market for AI data centers is projected to grow substantially, with co-packaged optics (CPO) alone expected to reach over $1 billion by 2034 from $95 million in 2025, according to industry estimates [20].

But this demand concentration creates fragility. If hyperscalers — Meta, Microsoft, Google, Amazon — slow their GPU cluster buildouts due to economic conditions, shifting AI workload patterns, or simply a pause in capital expenditure cycles, both Credo's existing AEC business and its newly acquired optical portfolio would be affected simultaneously. The deal's return on investment assumes the current AI infrastructure buildout continues at or near its present pace through at least fiscal 2027.

A second technology risk involves the Ultra Ethernet Consortium (UEC), which released its 1.0.1 specification in mid-2025 with over 100 member companies [21]. UEC aims to make standard Ethernet competitive with InfiniBand for AI back-end networks by adding RDMA, advanced flow control, and ultra-low tail latency while maintaining compatibility with existing Ethernet optics [21]. If UEC-based architectures become the dominant interconnect standard, they could shift purchasing decisions in ways that favor different optical component vendors or different integration approaches. However, UEC specifications rely on the same IEEE 802.3 Ethernet physical layer and optics [22], which means DustPhotonics' transceiver-level silicon photonics would likely remain relevant regardless of which higher-layer protocol prevails.

The Overpayment Question

The steelman case that $750 million is too much runs as follows: DustPhotonics is a startup with no disclosed recurring revenue, approximately 70 employees, and a product portfolio that, while technically advanced, has not yet been proven at the scale needed to generate the $500 million combined optical revenue target Credo has set for FY2027. The implied revenue multiple, even on optimistic forward projections, exceeds what larger, more established optical companies have commanded in comparable transactions.

Historical precedent offers mixed signals. Cisco's Acacia acquisition, initially priced at $2.6 billion in 2019, closed at $4.5 billion in 2021 after a protracted legal dispute [10][11]. Acacia had approximately $400 million in annual revenue at the time of the original bid [10], making the final price roughly 11x revenue. The II-VI/Coherent merger valued Coherent at approximately $7 billion against roughly $1.4 billion in annual revenue [12], or about 5x. By comparison, paying $750 million (base) for a company generating perhaps $50 million to $100 million in revenue — figures that are speculative given the lack of disclosure — would imply a multiple of 7.5x to 15x.

The counterargument is that DustPhotonics brings something acquisitions of mature companies do not: a technology roadmap extending to 3.2 Tbps and beyond, a team with deep photonics expertise that would take years to recruit, and existing relationships with hyperscaler customers [19][15]. In a market where NVIDIA alone committed $4 billion to photonics investments in a single quarter [13], $750 million for a focused silicon photonics capability with near-term revenue potential could prove inexpensive in hindsight.

Competitive Implications

The acquisition most directly affects Marvell Technology, which has positioned itself as a key supplier of PAM4 optics and custom silicon for cloud customers. Marvell's "NVLink Fusion" partnership with NVIDIA integrates optical engines into custom AI accelerators [13], an approach that competes with Credo's vision of a vertically integrated connectivity stack. A combined Credo-DustPhotonics could offer hyperscalers an alternative optical DSP and PIC platform outside the NVIDIA-Marvell ecosystem.

Coherent Corp (formed from the 2022 II-VI merger) and Lumentum Holdings are less directly threatened, as both have secured strategic investments from NVIDIA and serve broader markets including telecom, industrial, and defense [13][14]. Coherent reported $6.1 billion in revenue for fiscal 2024, with datacom and telecom representing about 40% of sales [12]. Their scale and diversification provide insulation that a focused company like Credo-DustPhotonics lacks.

Broadcom, dominant in networking switch silicon with over $20 billion in AI semiconductor revenue, represents the largest competitive force in the broader ecosystem [13]. Its position in both switch ASICs and optical DSPs gives it an integration advantage that Credo's acquisition only partially addresses.

Regulatory and Geopolitical Considerations

The acquisition involves an Israeli-headquartered company with international investors, which raises questions about regulatory review. DustPhotonics' investor base includes European venture capital firms (Exor Ventures, based in the Netherlands) and Israeli investment groups [8][9]. The company's technology — silicon photonics PICs for data center transceivers — falls within the scope of U.S. export control regulations governing advanced photonic integrated circuits.

While the transaction announcement references "customary closing conditions and regulatory approvals" [1], it does not specifically mention CFIUS (Committee on Foreign Investment in the United States) review. Credo itself is incorporated in the Cayman Islands and headquartered in San Jose, California [5], a corporate structure that may add complexity to the regulatory analysis. Given DustPhotonics' Israeli operations and its technology's relevance to AI infrastructure, the deal will likely face at least routine scrutiny under current U.S. trade and technology transfer frameworks, though significant regulatory obstacles appear unlikely given the allied relationship between the U.S. and Israel.

Export control implications may prove more consequential over time. As U.S. restrictions on advanced semiconductor exports to China and other countries continue to tighten, the ability to ship DustPhotonics' most advanced PICs to certain markets could be constrained, limiting the combined company's total addressable market.

What Comes Next

The market has signaled approval of the deal, with Jefferies initiating coverage of Credo at a Buy rating with a $175 price target and Mizuho maintaining its Outperform rating with a $200 target [23][24]. Both firms cite Credo's AI infrastructure positioning as the primary thesis. The transaction is expected to be accretive to adjusted earnings per share in fiscal 2027 [4].

The more meaningful test will come over the next 12 to 18 months, as Credo integrates DustPhotonics' 70-person team, scales production of 800G and 1.6T silicon photonics PICs, and demonstrates that its vertically integrated connectivity stack can win designs against Marvell, Broadcom, and the NVIDIA-backed optical supply chain. The $500 million optical revenue target for fiscal 2027 will serve as the clearest benchmark of whether $750 million was a price well paid — or the cost of a ticket to a race Credo cannot afford to lose.