Marvell's $540M XConn Acquisition Signals the Next Phase of AI Infrastructure

Marvell Technology committed $540 million to acquire XConn Technologies on January 6, 2026, marking the largest acquisition in the CXL switching silicon market to date [1]. The deal arrives as AI data centers face a structural crisis: memory bandwidth has become the primary bottleneck limiting GPU cluster performance, and traditional interconnects cannot keep pace with the demands of models exceeding 100 billion parameters [2]. With CXL 4.0 enabling 100+ terabyte memory pools and UALink 1.0 promising an open alternative to NVIDIA's proprietary NVLink, the interconnect layer has emerged as the critical infrastructure battleground for 2026 and beyond [3][4].

TL;DR

• Marvell acquires XConn for $540 million (60% cash, 40% stock), gaining hybrid PCIe/CXL switching silicon leadership [1][5].
• CXL 4.0 enables 100+ TB memory pools with 1.5 TB/s bandwidth across multiple racks, addressing AI's memory wall [6][7].
• UALink 1.0 delivers 200 Gb/s per lane for up to 1,024 accelerators, challenging NVIDIA's NVLink dominance [8][9].
• The hybrid switch market will reach $2.2 billion by 2026, growing at 12.3% CAGR [10].
• PCIe Gen6 and CXL 3.1 products begin shipping mid-2026, forcing infrastructure upgrades across AI deployments [11][12].

The Memory Wall Crisis Driving Interconnect Investment

AI infrastructure has hit a structural wall. Memory bandwidth, packaging interconnects, and thermal management now constrain performance more than raw GPU compute power [13]. SK Hynix CFO confirmed the company has "already sold out our entire 2026 HBM supply," while Micron reports high-bandwidth memory capacity remains fully booked through calendar year 2026 [14][15].

The numbers paint a stark picture of the bottleneck:

Constraint	Status	Impact
HBM Supply	Sold out through 2026	$100B TAM projected by 2028 [16]
CoWoS Packaging	"Very tight" per TSMC CEO	Limits GPU production [17]
Memory Prices	50% increase projected through Q2 2026	Infrastructure cost escalation [18]
DDR5 Server Memory	30-40% price increase Q4 2025	Doubling possible by 2026 [19]

Traditional interconnects compound the problem. A 70B parameter model with 128K context and batch size 32 can require 150+ GB for KV cache alone [20]. Moving data between accelerators at sufficient speed requires interconnects operating at terabytes per second.

"The true bottlenecks are no longer GPUs themselves, but memory bandwidth, packaging interconnects, thermal management, and power supply," notes analysis from Fusion Worldwide [17].

CXL memory pooling provides one solution: storing KV cache in pooled CXL memory while keeping hot layers in GPU VRAM [20]. Achieving 3.8x speedup compared to 200G RDMA and 6.5x speedup versus 100G RDMA, CXL dramatically reduces time-to-first-token (TTFT) for inference workloads [21].

Why Marvell Paid $540 Million for XConn

XConn Technologies holds a unique position in the interconnect market: the company developed the industry's first hybrid switch supporting both CXL and PCIe on a single chip [1][22]. Marvell's acquisition targets three strategic capabilities:

Production-Ready Technology Stack

XConn delivers products across multiple generations:

Product	Standard	Status
Current switches	PCIe 5.0 / CXL 2.0	Production shipping [22]
Apollo 2	PCIe 6.2 / CXL 3.1	Sampling (launched March 2025) [23]
Next-gen	PCIe 7.0 / CXL 4.0	Development [24]

The Apollo 2 hybrid switch integrates CXL 3.1 and PCIe Gen 6.2 on a single chip, offering support for the latest standards as they enter production [23].

Timing Advantages

The acquisition closes in early 2026, positioning Marvell to capture the PCIe Gen6 transition cycle [5]. PCIe Gen6 doubles bandwidth to 64 Gbps per lane but halves signal integrity distance, forcing server designers to deploy retimers on nearly every lane [25]. Every server shipping with next-generation accelerators will require this silicon.

2026 also marks early adoption of CXL memory pooling architectures, requiring connectivity modules that let processors "borrow" memory from adjacent devices [25]. XConn's hybrid approach addresses both requirements simultaneously.

Deal Structure

Marvell structures the transaction as approximately 60% cash and 40% stock, valued at $540 million total [1][5]. The mixed consideration signals confidence in long-term integration while managing cash outflow.

Matt Murphy, Marvell's CEO, characterized the strategic rationale: "XConn is the innovation leader in next-generation interconnect technology for high-performance computing and AI applications" [1].

CXL 4.0: Memory Pooling at Unprecedented Scale

The CXL Consortium released CXL 4.0 on November 18, 2025, doubling bandwidth to 128 GT/s with PCIe 7.0 integration [6][26]. The specification introduces capabilities that fundamentally change how AI infrastructure architects design memory systems.

Core Technical Advances

Feature	CXL 3.x	CXL 4.0
Bandwidth	64 GT/s	128 GT/s [6]
PCIe Base	PCIe 6.0	PCIe 7.0 [26]
Bundled Port Bandwidth	N/A	1.5 TB/s [7]
Memory Pool Scale	Single rack	Multi-rack (100+ TB) [27]

CXL 4.0 introduces Bundled Ports, allowing hosts and devices to aggregate multiple physical ports into single logical attachments [26]. A single bundled connection can deliver 1.5 TB/s bandwidth while maintaining a simplified software model [7].

Latency Characteristics

CXL provides memory-semantic access with latency in the 200-500 nanosecond range [28]. For comparison:

Technology	Typical Latency
Local DRAM	~100 ns
CXL Memory	200-500 ns [28]
NVMe Storage	~100 microseconds [28]
Storage-based Sharing	>10 milliseconds [28]

The 200-500 ns latency enables dynamic, fine-grained memory sharing across compute nodes that storage-based approaches cannot match [28].

Infrastructure Impact

CXL memory pooling has reduced hyperscaler total cost of ownership by an estimated 15-20% for memory-intensive workloads [29]. The technology addresses memory stranding by allowing unused capacity on one server to serve workloads on another.

AI inference workloads requiring hundreds of terabytes can now access shared memory pools across racks with cache coherency [27]. The multi-rack capability represents a fundamental shift from the single-server memory architecture that has dominated data center design for decades.

Deployment Timeline

Phase	Timeline	Capability
CXL 3.1 silicon sampling	H1 2026 [12]	PCIe 6.0 speeds, per-rack pooling
CXL 4.0 product sampling	Late 2026 [7]	128 GT/s, multi-rack
Multi-rack production	2026-2027 [30]	100+ TB pools, full disaggregation

AMD announced the Versal Premium Series Gen 2 as the first FPGA platform supporting CXL 3.1 and PCIe Gen6, with silicon samples expected by early 2026 and production units by mid-2026 [12].

UALink 1.0: The Open Alternative to NVLink

The Ultra Accelerator Link Consortium released UALink 1.0 on April 8, 2025, establishing an open standard for GPU/accelerator interconnects that challenges NVIDIA's proprietary NVLink [8][31]. The consortium includes AMD, Intel, Google, Microsoft, Meta, Broadcom, Cisco, HPE, and AWS, with Apple and Alibaba Cloud joining at board level in January 2025 [32][33].

Technical Specifications

UALink 1.0 delivers specifications competitive with NVIDIA's current NVLink offerings:

Specification	UALink 1.0	NVLink 4.0	NVLink 5.0
Per-Lane Bandwidth	200 Gb/s [8]	900 GB/s aggregate [34]	2,538 GB/s [34]
Max Accelerators in Pod	1,024 [9]	256 theoretical, 8 commercial [35]	576 theoretical, 72 commercial [35]
Consortium	Open standard [31]	NVIDIA proprietary	NVIDIA proprietary

A group of four lanes constitutes a "Station," offering maximum bandwidth of 800 Gbps bidirectional [36]. System designers can scale the number of accelerators and bandwidth allocated to each accelerator independently [36].

Competitive Positioning

UALink combines elements from PCI-Express, AMD's Infinity Fabric, and modified Ethernet SerDes to create a purpose-built interconnect for accelerator memory fabrics [37]. The specification achieves "the same raw speed as Ethernet with the latency of PCIe switches" according to consortium materials [38].

The security feature UALinkSec provides data confidentiality and optional data integrity including replay protection, supporting encryption and authentication across all protocol channels [39].

Hardware Timeline

UALink 1.0 hardware enters production in the 2026-2027 window [40]. AMD and Intel will ship accelerators supporting the standard, with Astera Labs and Broadcom delivering compatible switches [40].

Upscale AI targets Q4 2026 for scale-up UALink switches [41]. Korean startup Panmnesia announced sample availability of PCIe 6.0/CXL 3.2 Fabric Switch implementing port-based routing for CXL fabrics [42].

The Three-Fabric Architecture: How NVLink, CXL, and UALink Coexist

Modern AI infrastructure increasingly requires all three interconnect fabrics operating simultaneously, each serving distinct functions within the cluster [43][44].

Fabric Roles

Fabric	Primary Function	Latency Profile	Multi-Vendor
NVLink	GPU-to-GPU (NVIDIA only)	Higher, bandwidth-optimized	No [45]
UALink	Accelerator-to-accelerator	Higher, bandwidth-optimized	Yes [37]
CXL	CPU-Memory coherency, pooling	Lower (200-500 ns)	Yes [28]

CXL uses PCIe SerDes, resulting in lower error rates and lower latency with commensurate lower bandwidth [44]. NVLink and UALink utilize Ethernet-style SerDes, trading higher error rates and latency for significantly higher bandwidth [44].

Convergence Path

CXL addresses memory-capacity expansion and coherent data-sharing between hosts and accelerators [46]. UALink and NVLink (collectively termed "XLink" in industry discussions) provide direct, point-to-point connections optimized for accelerator-to-accelerator data exchanges [46].

Future architectures will likely deploy CXL for memory pooling and sharing between hosts, with remote scale-out over UALink and UltraEthernet fabrics [44]. Switches supporting both CXL and UALink represent the likely consolidation point [44].

Marvell's XConn acquisition directly targets building silicon for these converged switch architectures.

Infrastructure Implications for 2026 Deployments

Organizations planning AI infrastructure deployments face critical decisions as interconnect technologies mature. The transition requires coordinating multiple upgrade cycles simultaneously.

Power and Cooling Considerations

Next-generation interconnects consume significant power at the switch and retimer level. PCIe Gen6's reduced signal distance forces additional active components into every server design [25].

Component	Power Impact
PCIe Gen6 Retimers	Required on most lanes [25]
CXL Switches	New power budget category
Bundled Port Aggregation	Multiplied port power

Planning Horizon

Infrastructure teams must align multiple technology transitions:

Technology	Production Availability	Planning Implication
PCIe 6.0	Mid-2026 [12]	Server refresh required
CXL 3.1	Mid-2026 [12]	Switch infrastructure upgrade
UALink 1.0	Late 2026-2027 [40]	Accelerator platform decision
CXL 4.0	Late 2026-2027 [7]	Multi-rack architecture option

Vendor Lock-in Considerations

NVIDIA's NVLink remains proprietary and tightly coupled to NVIDIA hardware [45]. Organizations deploying non-NVIDIA accelerators or seeking multi-vendor flexibility should evaluate UALink-compatible hardware entering production in 2026-2027 [40].

CXL offers the broadest vendor ecosystem, with AMD, Intel, Samsung, SK Hynix, Micron, and dozens of smaller vendors shipping compatible products [47].

The Introl Advantage: Deploying Complex Interconnect Infrastructure

Deploying these interconnect technologies requires specialized expertise that extends beyond traditional server installation. The cabling, switch configuration, and topology design for CXL memory pools and UALink fabrics demand precise execution at scale.

Introl maintains 550 field engineers specialized in high-performance computing deployments across 257 global locations [48]. GPU cluster installations increasingly require integrating CXL switches, managing retimer placement, and validating end-to-end fabric performance before production handoff.

For organizations scaling from dozens to thousands of accelerators, professional deployment teams understand the nuances of next-generation interconnects. Fiber optic connections spanning over 40,000 miles require careful attention to signal integrity requirements that PCIe Gen6 and CXL 3.1 demand [48][49].

Key Takeaways by Role

Infrastructure Planners

• Budget for PCIe Gen6 server refresh in 2026; retimers add component cost and power
• Evaluate CXL switch vendors now; lead times will extend as demand increases
• Plan rack layouts for multi-rack CXL pooling if memory-intensive AI inference workloads dominate

Operations Teams

• Develop CXL fabric monitoring capabilities before deployment
• Train staff on UALink topology configuration for non-NVIDIA accelerator environments
• Establish signal integrity testing procedures for PCIe Gen6 distances

Strategic Decision Makers

• The Marvell-XConn acquisition signals consolidation; expect fewer, larger interconnect vendors
• UALink provides optionality against NVIDIA lock-in for accelerator purchases
• CXL memory pooling can reduce TCO 15-20% for appropriate workloads; validate against your specific applications

Looking Ahead: The Interconnect Imperative

The interconnect layer has transformed from passive infrastructure to active differentiator for AI deployments. Marvell's $540 million bet on XConn reflects the strategic importance of controlling switching silicon as memory and accelerator fabrics converge.

Organizations deploying AI infrastructure in 2026 and beyond must treat interconnect selection as a first-order architectural decision. The choice between proprietary NVLink, open UALink, and memory-focused CXL will shape flexibility, cost structure, and performance for years after installation.

The winners in the next phase of AI infrastructure buildout will master all three fabrics simultaneously. Those who treat interconnects as commodity commoditized components will find their GPU investments underperforming as memory walls and bandwidth constraints limit what their accelerators can achieve.

---

References

[1] Marvell Technology. "Marvell to Acquire XConn Technologies, Expanding Leadership in AI Data Center Connectivity." Marvell Investor Relations. January 6, 2026. https://investor.marvell.com/news-events/press-releases/detail/1004/marvell-to-acquire-xconn-technologies-expanding-leadership-in-ai-data-center-connectivity

[2] Keysight. "Key Challenges in Scaling AI Data Center Clusters." Keysight Blogs. February 11, 2025. https://www.keysight.com/blogs/en/inds/2025/2/11/key-challenges-in-scaling-ai-data-center-clusters

[3] CXL Consortium. "CXL 4.0 Specification Release." November 18, 2025. https://computeexpresslink.org/

[4] UALink Consortium. "UALink 200G 1.0 Specification Release." April 8, 2025. https://ualinkconsortium.org/

[5] Yahoo Finance. "Marvell to Acquire XConn Technologies, Expanding Leadership in AI Data Center Connectivity." January 6, 2026. https://finance.yahoo.com/news/marvell-acquire-xconn-technologies-expanding-140000224.html

[6] Blocks and Files. "CXL 4.0 doubles bandwidth and stretches memory pooling to multi-rack setups." November 24, 2025. https://blocksandfiles.com/2025/11/24/cxl-4/

[7] Introl. "CXL 4.0 and the Interconnect Wars: How AI Memory Is Reshaping Data Center Architecture." December 2025. https://introl.com/blog/cxl-4-0-specification-interconnect-wars-december-2025

[8] The Register. "UALink debuts its first AI interconnect spec." April 8, 2025. https://www.theregister.com/2025/04/08/ualink_200g_version_1/

[9] Data Center Dynamics. "UALink Consortium releases 200G 1.0 specification for AI accelerator interconnects." April 2025. https://www.datacenterdynamics.com/en/news/ualink-consortium-releases-200g-10-specification-for-ai-accelerator-interconnects/

[10] Grand View Research. "Hybrid Switch Market Report." 2025. Via StockTitan analysis. https://www.stocktitan.net/news/MRVL/marvell-to-acquire-x-conn-technologies-expanding-leadership-in-ai-72p1mhcm3x06.html

[11] Network Computing. "Choosing the Right Interconnect for Tomorrow's AI Applications." 2025. https://www.networkcomputing.com/data-center-networking/choosing-the-right-interconnect-for-tomorrow-s-ai-applications

[12] All About Circuits. "AMD First to Release FPGA Devices With CXL 3.1 and PCIe Gen6." 2025. https://www.allaboutcircuits.com/news/amd-first-release-fpga-devices-with-cxl-3.1-pcie-gen6/

[13] AInvest. "The Critical AI Memory Infrastructure Bottleneck and Its Investment Implications." December 2025. https://www.ainvest.com/news/critical-ai-memory-infrastructure-bottleneck-investment-implications-2512/

[14] Medium. "Memory Supercycle: How AI's HBM Hunger Is Squeezing DRAM." December 2025. https://medium.com/@Elongated_musk/memory-supercycle-how-ais-hbm-hunger-is-squeezing-dram-and-what-to-own-79c316f89586

[15] Introl. "The AI Memory Supercycle: How HBM Became AI's Most Critical Bottleneck." 2026. https://introl.com/blog/ai-memory-supercycle-hbm-2026

[16] Medium. "The Next Five Years of Memory, And Why It Will Decide AI's Pace." 2025. https://medium.com/@Elongated_musk/the-next-five-years-of-memory-and-why-it-will-decide-ais-pace-27c4318fe963

[17] Fusion Worldwide. "Inside the AI Bottleneck: CoWoS, HBM, and 2-3nm Capacity Constraints Through 2027." 2025. https://www.fusionww.com/insights/blog/inside-the-ai-bottleneck-cowos-hbm-and-2-3nm-capacity-constraints-through-2027

[18] Counterpoint Research. Via Catalyst Data Solutions. "Memory Shortage in 2026: How AI Demand Is Reshaping Supply?" https://www.catalystdatasolutionsinc.com/the-lab/ddr5-memory-shortage-2026/

[19] Catalyst Data Solutions. "Memory Shortage in 2026: How AI Demand Is Reshaping Supply?" 2026. https://www.catalystdatasolutionsinc.com/the-lab/ddr5-memory-shortage-2026/

[20] Medium. "CXL: The Secret Weapon to Solving the AI Memory Wall." January 2026. https://medium.com/@tanmaysorte25/cxl-the-secret-weapon-to-solving-the-ai-memory-wall-c22f93e8547d

[21] CXL Consortium. "Overcoming the AI Memory Wall: How CXL Memory Pooling Powers the Next Leap in Scalable AI Computing." 2025. https://computeexpresslink.org/blog/overcoming-the-ai-memory-wall-how-cxl-memory-pooling-powers-the-next-leap-in-scalable-ai-computing-4267/

[22] Data Center Dynamics. "Marvell acquires PCIe and CXL switch provider XConn Technologies for $540m." January 2026. https://www.datacenterdynamics.com/en/news/marvell-acquires-pcie-and-cxl-switch-provider-xconn-technologies-for-540m/

[23] XConn Technologies. "Apollo 2 Hybrid Switch Launch." March 2025. Via Marvell acquisition materials.

[24] CXL Consortium. "CXL Roadmap." 2025. Via VideoCardz. https://videocardz.com/newz/cxl-4-0-spec-moves-to-pcie-7-0-doubles-bandwidth-over-cxl-3-0

[25] Network Computing. "The transition to PCIe Gen 6 is the critical driver for 2026." 2025. https://www.networkcomputing.com/data-center-networking/choosing-the-right-interconnect-for-tomorrow-s-ai-applications

[26] VideoCardz. "CXL 4.0 spec moves to PCIe 7.0, doubles bandwidth over CXL 3.0." November 2025. https://videocardz.com/newz/cxl-4-0-spec-moves-to-pcie-7-0-doubles-bandwidth-over-cxl-3-0

[27] Introl. "CXL 4.0 Infrastructure Planning Guide: Memory Pooling for AI at Scale." 2025. https://introl.com/blog/cxl-4-0-infrastructure-planning-guide-memory-pooling-2025

[28] CXL Consortium. "How CXL Transforms Server Memory Infrastructure." October 2025. https://computeexpresslink.org/wp-content/uploads/2025/10/CXL_Q3-2025-Webinar_FINAL.pdf

[29] KAD. "CXL Goes Mainstream: The Memory Fabric Era in 2026." 2026. https://www.kad8.com/hardware/cxl-opens-a-new-era-of-memory-expansion/

[30] GIGABYTE. "Revolutionizing the AI Factory: The Rise of CXL Memory Pooling." 2025. https://www.gigabyte.com/Article/revolutionizing-the-ai-factory-the-rise-of-cxl-memory-pooling

[31] Network World. "UALink releases inaugural GPU interconnect specification." April 2025. https://www.networkworld.com/article/3957541/ualink-releases-inaugural-gpu-interconnect-specification.html

[32] Blocks and Files. "The Ultra Accelerator Link Consortium has released its first spec." April 9, 2025. https://blocksandfiles.com/2025/04/09/the-ultra-accelerator-link-consortium-has-released-its-first-spec/

[33] The Next Platform. "Key Hyperscalers And Chip Makers Gang Up On Nvidia's NVSwitch Interconnect." May 30, 2024. https://www.nextplatform.com/2024/05/30/key-hyperscalers-and-chip-makers-gang-up-on-nvidias-nvswitch-interconnect/

[34] LoveChip. "UALink vs NVLink: What Is the Difference?" 2025. https://www.lovechip.com/blog/ualink-vs-nvlink-what-is-the-difference-

[35] The Next Platform. "UALink Fires First GPU Interconnect Salvo At Nvidia NVSwitch." April 8, 2025. https://www.nextplatform.com/2025/04/08/ualink-fires-first-gpu-interconnect-salvo-at-nvidia-nvswitch/

[36] Converge Digest. "UALink 1.0 Released for Low-Latency Scale-Up AI Accelerators." 2025. https://convergedigest.com/ualink-1-0-released-for-low-latency-scale-up-ai-accelerators/

[37] NAND Research. "Research Note: UALink Consortium Releases UALink 1.0." 2025. https://nand-research.com/research-note-ualink-consortium-releases-ualink-1-0/

[38] Astera Labs. "Building the Case for UALink: A Dedicated Scale-Up Memory Semantic Fabric." 2025. https://www.asteralabs.com/building-the-case-for-ualink-a-dedicated-scale-up-memory-semantic-fabric/

[39] UALink Consortium. "UALink 1.0 Specification." April 2025. Via Data Center Dynamics. https://www.datacenterdynamics.com/en/news/ualink-consortium-releases-200g-10-specification-for-ai-accelerator-interconnects/

[40] Futuriom. "UALink Offers Fresh Options for AI Networking." April 2025. https://www.futuriom.com/articles/news/ualink-spec-offers-fresh-scale-up-options/2025/04

[41] HPCwire. "Upscale AI Eyes Late 2026 for Scale-Up UALink Switch." December 2, 2025. https://www.hpcwire.com/2025/12/02/upscale-ai-eyes-late-2026-for-scale-up-ualink-switch/

[42] Blocks and Files. "Panmnesia pushes unified memory and interconnect design for AI superclusters." July 18, 2025. https://blocksandfiles.com/2025/07/18/panmnesia-cxl-over-xlink-ai-supercluster-architecture/

[43] Clussys. "Towards Tomorrow's AI Networking: RDMA and IP over CXL Fabric and More." June 18, 2024. https://clussys.github.io/blogs/2024-06-18-ai-networking

[44] Semi Engineering. "CXL Thriving As Memory Link." 2025. https://semiengineering.com/cxl-thriving-as-memory-link/

[45] ServeTheHome. "UALink will be the NVLink Standard Backed by AMD Intel Broadcom Cisco and More." 2024. https://www.servethehome.com/ualink-will-be-the-nvlink-standard-backed-by-amd-intel-broadcom-cisco-and-more/

[46] SlideShare. "Memory over Fabrics: An Open Journey from CXL to UALink in AI Infrastructure." 2025. https://www.slideshare.net/slideshow/memory-over-fabrics-an-open-journey-from-cxl-to-ualink-in-ai-infrastructure/276631394

[47] Wikipedia. "Compute Express Link." https://en.wikipedia.org/wiki/Compute_Express_Link

[48] Introl. "Company Overview." https://introl.com/coverage-area

[49] Rivosinc. "Ultra Ethernet Specification 1.0 – A Game Changer for AI Networking." 2025. https://www.rivosinc.com/resources/blog/ultra-ethernet-specification-1-0-a-game-changer-for-ai-networking

[50] SemiAnalysis. "The New AI Networks | Ultra Ethernet UEC | UALink vs Broadcom Scale Up Ethernet SUE." June 11, 2025. https://semianalysis.com/2025/06/11/the-new-ai-networks-ultra-ethernet-uec-ualink-vs-broadcom-scale-up-ethernet-sue/

[51] APNIC Blog. "Scale-up fabrics." June 3, 2025. https://blog.apnic.net/2025/06/03/scale-up-fabrics/

[52] EE Times. "DRAM Cannot Keep Up With AI Demand." 2025. https://www.eetimes.com/dram-cannot-keep-up-with-ai-demand/

[53] EE Times Asia. "Memory Becoming Chip Industry's Next Bottleneck Amid Strong AI Demand." 2025. https://www.eetasia.com/memory-becoming-chip-industrys-next-bottleneck-amid-strong-ai-demand/

[54] IAEME. "The Evolution of PCI Express: From Gen1 to Gen6 and Beyond." International Journal of Computer Engineering and Technology. 2025. https://iaeme.com/Home/article_id/IJCET_16_01_153

[55] ExoSwan. "Top AI Infrastructure Stocks 2026: Data Center Picks & Shovels." 2026. https://exoswan.com/ai-infrastructure-stocks