Where is the Next AI investment Opportunity?
Who will benefit next as AI investment deepens and moves up from infrastructure to application layer?
GREY RHINO
The AI investment is a boon to many businesses. The whole ecosystem evolves and expands as AI investment enters different phases. Back in 2023-2024 when generative AI was at the LLM training stage, GPU makers like Nvidia (NVDA) were the main beneficiaries of demand from AI Hyperscalers. CPU makers like Intel got the blow for missing the GPU boat. Fast forwarding today, when Gen AI moves investment from training to inference, CPU suddenly becomes a hot commodity. Intel (INTL) rode the wave to historical high, making its investors very happy for the last 6 months. AMD benefits from its business in both GPU and CPU, and is now trading at a lofty valuation.
Such is the effect of “rising tide raising all boats.” More accurately, it is “riding tide raising different boats at different junctures of the river bed.” As AI enters different phases, the demand shifts, sometimes creating choke points that benefit specific groups of suppliers and partners. So where are we and what sectors have benefited? Most importantly, who will benefit next as AI investment deepens and moves up from infrastructure to application layer?
The first phase of Gen AI investment benefited silicon makers. Nvidia is the best example. Yet, as AI computational power increases and more AI chip makers enter the race, the demand showed new paths. One is the CPU path for inference computation that was already mentioned. The other path is hyperscalers’ demand for customized AI ASICs designed to fit their own LLM models and computational needs. This demand drives up sales for companies like Broadcom (AVGO) and Marvell (MRVL). Even IC design firms like ARM (ARM) also got a boost as demand for such ICs is booming.
Then these hyperscalers, confident of their GenAI model’s capability, began to build AI infrastructure and more specifically data centers. This new phase, which is in full swing now, benefits a slew of partners and suppliers, including the following types:
Server rack assemblers: these companies such as Dell (DELL), Super Micro Computer (SMCI), HP (HPE), build GPU and CPU server racks that can be installed in data centers. This group just grabbed the spotlight today with DELL’s 40% rise after earning.
Data center builders: Hyperscalers are investing in building their own data centers but also contracting 3rd party partners to build data centers for them to lease. Oracle (ORCL)is well known for raising a significant amount of debt to fulfill its build obligation of data centers to these hyperscalers. Nebuis (NBIS) and CoreWeave (CRWV) are the other two recent high flyers.
Energy companies, particularly mid-stream and downstream companies. AI data centers consume a significant amount of energy, which boost energy suppliers that either transport oil and gas, or produce electricities. Best examples are Kinder Morgan (KMI), DT MIdstream (DTM), Vistra (VST), and Talen Energy (TLN).
Networking companies: in a data center where hundreds if not thousands of GPU- or CPU-based servers are running inside, massive traffic on and between these servers need smarter routers, switches, and controllers. AI adoption is creating a large network refresh cycle that demands these faster data center networking equipment and software. Networking companies like Cisco (CSCO) and Marvell (MRVL) saw demand for their hardware and software skyrocket as well.
Fiber/Optic: When thousands of GPUs are clustered together, data must be converted from electrical signals to light via optical transceivers to prevent latency bottlenecks. Optical chip companies or optical switch makers like Corning (GLW), Lumentum (LITE), Coherent (COHR) have become the latest market darlings.
Memory Suppliers: AI computation apparently needs a lot of DRAMs and storage. AI Demand is pushing prices of these memory products to unprecedented levels that DRAM companies like Micron (MU), SD Hynix (STH) , and Samsung, as well SSD/Storage names like Sandisk (SNDK), Western Digital (WDC), Seagate (STX), are reaching all time highs. What used to be a commodity category subjected to very cyclical demand cycles is now seeing consistent demand that outstrips their manufacturing capacity, giving them enormous pricing power and long backlogs.
Now, where is the next investment opportunity if you missed all the above? A lot of promises are placed on enterprise AI applications powered by GPUs/CPUs/TPUs, server racks, data centers, and memory chips, optical networking chips, and energy sources. Although the entire software sector was hammered for fear of AI slashing its utility and shaking the foundation of this sector’s SaaS revenue model, Snowflake (SNOW) recently demonstrated through its 1Q earning report that using generative AI can boost software revenue. The growth comes from two sources. First, Snowflake’s customers realize that in order to train AI models, build AI agents, or run generative queries, companies must first organize massive troves of clean enterprise data to prepare for AI-enabled efficiency. This leads to direct investment in Snowflake’s data warehousing platform. The second source comes from Snowflake’s own first-party AI capabilities–its coding agent, CoCo. CoCo allows non-technical business employees to ask natural language questions about the complex data stored in Snowflake. Because Snowflake charges per query, this democratization of data caused an explosive expansion of billable computer workloads across more than 13,600 accounts using Snowflake AI capabilities.
Snowflake’s early success indicates that the next AI investment opportunities will come from data warehousing/data lake companies that can normalize/standardize complex enterprise data before AI agents can work magic. Database companies with first-party AI capabilities can be the next winners. Snowflake's earnings marked a critical turning point, proving that the software layer is now successfully capturing the AI capital expenditure wave by letting enterprises turn raw data into productivity gains.
Trivia Economics © 2024-2028
Trivia Economics broadens/deepens your understanding of economics and investment opportunities.
Coming Soon
Subscribe to our newsletter and never miss a trivia.
We care about your data in our privacy policy.