Bank of America predicts that the AI infrastructure prosperity will be "stronger and longer"! By 2030, the scale will reach 1.7 trillion, with NVIDIA Corporation (NVDA.US) and AMD (AMD.US) still on the "preferred list".

When the market is still debating whether artificial intelligence is an expensive "tech bubble," Bank of America Securities pointed out in its latest research report that AI-related spending will not only not ebb, but will instead be "stronger and more enduring." In this report, NVIDIA Corporation (NVDA.US), AMD (AMD.US), and Broadcom Inc. (AVGO.US) continue to be listed as "preferred names" in the semiconductor sector, while Micron Technology, Inc. (MU.US) and Marvell Technology, Inc. (MRVL.US) have also made it to the core list. AI spending not only needs to be "stronger," but also "longer lasting" Given the capital expenditure outlook shown by major hyperscale cloud service providers such as Microsoft Corporation (MSFT.US), Amazon.com, Inc. (AMZN.US), Alphabet (GOOGL.US), and Meta Platforms (META.US) in their recent first quarter earnings reports, Bank of America predicts that by 2030, the market size of AI data center systems will soar from the previously predicted $1 trillion to an astonishing $1.7 trillion. Out of the potential market size of $1.7 trillion, around $1.2 trillion may come from AI accelerators, higher than the previous estimate of $1 trillion. This upward revision is attributed to the increase in shipments of custom ASICs by hyperscale cloud service providers, such as Alphabet Inc. Class C's TPU or Amazon.com, Inc.'s Trainium chip. The market size of data center CPUs may reach approximately $110 billion (above the previous estimate of $80 billion), while the AI networking market size may increase to around $316 billion (higher than the previous estimate of approximately $240 billion). Analysts explained, "Importantly, we note that with the extension of AI workload tail effects, compute/memory components are continuously diversifying. But we view this as complementary to the overall potential market size, and not as a replacement for existing workloads or components. For example, we expect CPUs (in their new standalone racks) to work in tandem with existing GPU-CPU compute racks, while SRAM-based ultra-low latency memory racks will coexist with HBM-based GPU racks." From "buying shovels" to "mining gold": 2026 will herald the ROI tipping point Over the past two years, Wall Street's investment logic in AI has primarily focused on "infrastructure rush" that is, tech giants have had to madly order GPUs from NVIDIA Corporation in order not to fall behind. However, this "burning cash for the future" pattern will always be questioned about profit prospects. Bank of America analysts believe that as AI unicorns like OpenAI and Anthropic are about to embark on the IPO process, the market will clearly see how these "top model manufacturers" will turn compute power into real commercial value. Once these startups prove their profit loop in the secondary market, the trillion-dollar expenditures from Microsoft Corporation, Amazon.com, Inc., Alphabet (parent company of Alphabet Inc. Class C), and Meta will have a rational and clear financial logic. It is no longer blindly "buying shovels," but has already found the "edge of the gold mine." Analysts particularly emphasize that 2026 will be a key point for the accelerated realization of AI sales and return on investment (ROI), while 2027 will need to verify whether the "token economy efficiency" (i.e., cost reduction of unit compute power) can be achieved as scheduled. However, beneath the fervor of expectations, Bank of America also retains rare calmness. The research report acknowledges that current supply chain bottlenecks may limit the short-term shipments of cutting-edge components, and the market always has doubts about the sustainability of capital expenditure. Lastly, Bank of America also raised the target prices of the following companies in the report: NVIDIA Corporation (from $300 to $320), AMD (from $450 to $500), Marvell Technology, Inc. (from $125 to $200), and Micron (from $500 to $950). The target price of Broadcom Inc. remains unchanged. Giving investors a "reality check" While Bank of America's bullish stance is certainly encouraging, as rational investors, several potential risks should be cautioned against: First is the risk of "IPO illusion." If the performance of OpenAI or Anthropic's IPO falls short of expectations, causing the valuation bubble in the primary market to burst, it may trigger a chain reaction, shaking the investment confidence of cloud vendors. Second is the "efficiency paradox." If algorithm optimization (such as model compression, inference acceleration) significantly reduces the compute power consumed by a unit AI task, the growth in hardware demand may not keep up with market expectations. Lastly is the "black swan" of the macro environment. The $1.7 trillion TAM forecast is built on the current relatively loose interest rate environment. If the Federal Reserve maintains high-interest rates in the long term, the capital costs of cloud vendors may rise, forcing them to cut long-term expenditure plans. However, in any case, Bank of America's report sends a clear signal: AI infrastructure investment is not a fleeting "event-driven" occurrence, but a "structural bull market" that will last at least until 2030. In this feast, NVIDIA Corporation is still the one "selling shovels," but AMD, Broadcom Inc., Marvell Technology, Inc., and Micron are all digging their own gold mines from different corners.