The artificial intelligence boom is creating a new infrastructure challenge as AI companies scale increasingly demanding models, and the bottleneck may no longer be limited to computing power.
Tema ETFs is betting that optical infrastructure will be an increasingly important part of the AI buildout, with Anthropic now accounting for roughly 15% of its Tema Photonics & Optical ETF‘s (NASDAQ:LAZR) holdings, making it one of the largest positions in the ETF.
Maurits Pot, founder and CEO of Tema ETFs, told Benzinga that Anthropic faces three core AI infrastructure bottlenecks.
"Anthropic faces 3 core AI infra bottlenecks – compute, memory, and optical," Pot said. "Anthropic benefits through AI infrastructure from the photonics innovations taking place that debottleneck optical."
The comment offers a more specific explanation for why an AI model developer has become a significant holding in a photonics-focused ETF. Rather than viewing Anthropic solely as an AI software company, TEMA sees its rapidly expanding infrastructure requirements as part of the investment case for the optical technologies needed to support AI at scale.
AI’s Optical Bottleneck
The explosive growth of AI workloads has driven massive investment in GPUs and other accelerators, but moving data between those processors is becoming an increasingly important challenge.
As AI systems scale, data centers need to move enormous volumes of information between GPUs, servers and networking infrastructure with minimal latency. Optical technologies can help address those bandwidth and connectivity demands, making photonics an increasingly important component of AI infrastructure.
For LAZR, that creates an investment opportunity that extends beyond semiconductor manufacturers and cloud companies.
"LAZR’s Anthropic exposure allows investors to both capture Anthropic’s growth as well as the downstream demand for photonics innovations around optical networking," Pot said.
That gives the ETF a two-layer AI exposure. Anthropic provides direct exposure to the potential growth of a leading AI model developer, while LAZR’s broader portfolio is positioned to benefit from the infrastructure spending required to support that growth.
Why Anthropic Belongs in a Photonics ETF
LAZR’s 15% weighting in Anthropic is notable because the AI developer isn’t a photonics or optical company at all. The investment instead reflects TEMA’s view that the infrastructure requirements of major AI companies can create demand throughout the optical technology ecosystem.
Anthropic’s growing compute requirements mean more pressure on data-center architecture, networking and connectivity. If those requirements continue to expand, companies developing technologies that improve how AI systems communicate could become increasingly important beneficiaries.
"Anthropic’s growing AI infrastructure needs and bottlenecks will benefit all the LAZR holdings, albeit in different ways," Pot said.
That broader exposure is central to the ETF’s thesis. Instead of trying to identify a single winner in AI infrastructure spending, LAZR seeks to capture multiple layers of the ecosystem, including the technologies that help move data through increasingly complex AI systems.
The Bigger AI Infrastructure Trade
The shift toward optical infrastructure comes as investors increasingly look beyond the traditional GPU-driven AI trade.
The first phase of the AI boom was dominated by demand for computing power. But as AI models become larger and data-center workloads more interconnected, bottlenecks can emerge elsewhere in the system.
That is where photonics could become strategically important.
For investors, LAZR’s Anthropic allocation therefore represents more than a bet on a potential blockbuster AI IPO. It is also a way for TEMA to express its view that AI infrastructure will increasingly depend on solving the optical bottleneck alongside compute and memory constraints.
With Anthropic now representing 15% of LAZR, TEMA is effectively tying the growth of one of the leading private AI companies to a broader bet on the technologies needed to make the next generation of AI infrastructure work.
Photo: PhotoGranary02 / Shutterstock
Login to comment