KEY HIGHLIGHTS

  • Bank of America raises Nvidia price target to $320 from $300, maintaining a Buy rating — implying further meaningful upside from current levels.
  • BofA upgrades its 2030 AI data centre TAM estimate to $1.7 trillion, up from $1.4 trillion — a $300 billion upward revision to the addressable market.
  • 2026 is expected to be a year of accelerating AI sales and return-on-Investment, with enterprise AI deployment moving from pilot to production at scale.
  • 2027 could prove a pivotal year as improved tokenomics and new architecture compute and memory systems ramp, potentially driving a second wave of efficiency gains.
  • Nvidia's dominant position in AI accelerators — holding an estimated 70–85% Market Share in data centre GPUs — places it at the centre of every major AI infrastructure build.
  • With a market cap above $3 trillion, Nvidia is now one of the most valuable companies in history — and the BofA revision suggests the growth story is far from over.

Bank of America has raised its price target on Nvidia to $320 from $300, reaffirming a Buy rating in a note that materially upgrades the firm's long-term outlook on the AI data centre market. The revision is not merely a routine price target nudge — it is underpinned by a significant upward revision to the total addressable market for AI infrastructure, a bullish read on the near-term Demand trajectory, and a forward-looking thesis about how the Economics of artificial intelligence will evolve through the second half of the decade. For investors trying to assess whether Nvidia's extraordinary run still has further to go, the BofA note provides a structured framework for thinking about the opportunity ahead.

The TAM Revision: $1.7 Trillion by 2030

 

The headline figure in the BofA update is the revised total addressable market estimate for AI data centre systems in calendar year 2030: $1.7 trillion, up from a prior estimate of $1.4 trillion. To put that number in context, the entire global semiconductor market — covering every chip, in every device, across every application — was approximately $580 billion in 2024. A single application vertical, AI data centre infrastructure, is now projected by one of Wall Street's most influential technology research teams to reach nearly three times that size within five years. The $300 billion upward revision reflects a more bullish assessment of enterprise AI adoption rates, the Capital-expenditure/">Capital Expenditure commitments being made by hyperscalers, and the expanding role of AI inference — running trained models to generate outputs — which is proving to be far more compute-intensive than the market anticipated even twelve months ago.

The TAM revision matters for Nvidia specifically because of the company's structural position within it. Nvidia's data centre segment — which includes its flagship H100, H200, and Blackwell GPU families, along with networking products from the Mellanox Acquisition and software through the CUDA ecosystem — is not a peripheral participant in the AI infrastructure market. It is, to a degree that would be remarkable in any industry, the market. The company's GPUs are the default compute substrate for Training and running large language models, image generation systems, scientific simulation workloads, and an expanding range of enterprise AI applications. A $300 billion expansion in the TAM, with Nvidia holding dominant share, translates directly into a material upgrade to long-term Revenue and Earnings expectations.

2026: Acceleration, Not Deceleration

 

A recurring concern among Nvidia sceptics has been the question of whether the current cycle of AI infrastructure investment is sustainable, or whether it represents a capital expenditure Bubble that will eventually deflate when returns Fail to materialise. BofA's note addresses this concern directly, describing 2026 as a year of accelerating AI sales and ROIs. This is a significant statement. It implies that the hyperscalers and enterprises investing in AI infrastructure — Microsoft, Google, Amazon, Meta, and a growing roster of sovereign AI initiatives and large corporates — are beginning to see tangible, measurable returns on the capital they have deployed. When infrastructure investment generates demonstrable returns, it creates a self-reinforcing cycle: returns justify further investment, further investment improves capabilities, improved capabilities unlock new use cases, new use cases drive further returns.

The enterprise AI deployment story is particularly important for the 2026 thesis. For much of 2023 and 2024, AI investment was concentrated in a small number of hyperscale players building foundational model infrastructure. The broadening of that investment into enterprise deployments — where companies are embedding AI into core Business processes, customer interactions, and product development workflows — represents a structural deepening of demand that is less susceptible to the kind of sharp Reversal that a single hyperscaler's capex reset might trigger. BofA's view that 2026 will see accelerating ROIs suggests this broadening is happening on a timeline that is ahead of the more conservative market consensus.

2027: The Tokenomics and Architecture Inflection

The most forward-looking element of the BofA thesis concerns 2027 and what the firm describes as improving tokenomics and the ramp of new architecture compute and memory systems. Tokenomics, in this context, refers to the cost efficiency of generating AI outputs — specifically, the cost per token produced by a large language model. Token costs have been falling rapidly as model architectures improve, inference hardware becomes more efficient, and software optimisation compounds over time. This deflationary dynamic is not a headwind for Nvidia — it is, counterintuitively, an accelerant. Lower token costs expand the universe of applications for which AI is economically viable, driving broader deployment and, in turn, higher aggregate compute demand even as per-unit costs decline. More applications requiring AI at lower cost per query means more total queries — and more total compute.

The reference to new architecture compute and memory systems points to the next generation of Nvidia hardware entering production ramp through 2026 and 2027. The Blackwell architecture, launched in 2024, represented a step-change in training and inference performance. Its successor, the Rubin architecture — already confirmed for 2026 delivery — is expected to deliver another significant leap in compute density and energy efficiency. Alongside compute, memory architecture is evolving rapidly: High Bandwidth Memory (HBM) generations are advancing, and the integration of compute and memory on the same package — a trend Nvidia is actively pursuing — has the potential to remove one of the key bottlenecks in current AI workloads. BofA's thesis implies that these hardware transitions will coincide with the enterprise deployment acceleration to create a powerful demand confluence in 2027.

Nvidia's Financial Momentum: The Numbers Behind the Thesis

The BofA price target revision does not exist in a vacuum — it reflects a business that has been executing at a level rarely seen in the history of large-cap technology. Nvidia's most recent full-year results for fiscal year 2025 (ending January 2025) showed total revenue of $130.5 billion, up 114% year-on-year. Data centre revenue alone reached $115.2 billion — a figure that, to contextualise, exceeds the entire annual revenue of many Fortune 100 companies. Net Income for the year was $72.9 billion, representing a net Margin of approximately 56%. Gross margins in the data centre segment have consistently run above 70%, reflecting the pricing power that accompanies technological Leadership and Supply constraints in a high-demand market.

For fiscal Q4 2025, Nvidia reported data centre revenue of $35.6 billion — up 93% year-on-year — driven by surging demand for Blackwell systems. The company guided fiscal Q1 2026 revenue to approximately $43 billion, ahead of analyst consensus, signalling no sign of demand deceleration entering the current fiscal year. Free Cash Flow generation has been extraordinary, enabling Nvidia to return capital to shareholders through Buybacks at scale while simultaneously investing in R&Amp;D, Manufacturing partnerships, and software ecosystem development. The CUDA software platform — which locks developers into Nvidia's hardware through years of optimised tooling, libraries, and frameworks — represents an additional layer of competitive moat that is not captured in hardware market share statistics alone.

Competitive Landscape: Challenges to the Throne

No investment thesis is complete without an honest assessment of competitive risk, and Nvidia faces genuine challenges across multiple dimensions. AMD's Instinct MI300 and MI325 series have gained traction in certain training workloads and are being adopted by hyperscalers looking to diversify their supply chains. Intel's Gaudi accelerators have found limited but growing adoption in cost-sensitive enterprise deployments. Custom silicon — the in-house AI chips being developed by Google (TPUs), Amazon (Trainium and Inferentia), Microsoft (Maia), and Meta (MTIA) — represents perhaps the most structurally significant long-term competitive pressure, as hyperscalers look to reduce their dependence on third-party hardware for certain high-Volume inference workloads.

These competitive pressures are real, and investors should not dismiss them. However, the historical record suggests that Nvidia's combination of hardware performance leadership, software ecosystem depth, and continuous architectural innovation has proven remarkably durable. The company has consistently stayed multiple generations ahead of its nearest GPU competitor, and the CUDA moat — representing billions of developer-hours of optimised code — is not easily replicated. Custom silicon, while growing, remains largely confined to specific, high-volume, well-defined workloads; the flexibility and general-purpose capability of Nvidia's GPUs retains significant advantage for the diverse, evolving, and often novel workloads that characterise cutting-edge AI development.

Valuation and the $320 Target: Is the Price Right?

Nvidia's Capitalisation/">Market Capitalisation above $3 trillion places it among the three most valuable companies in the world, alongside Apple and Microsoft. At this scale, the law of large numbers begins to exert pressure on growth expectations. The BofA price target of $320 implies a further appreciation of approximately 10–15% from the levels at which the note was published — a more modest implied upside than target revisions in earlier stages of the AI cycle, but still meaningful for a company of this capitalisation. The forward earnings multiple at which Nvidia trades — approximately 35–40 times fiscal year 2026 consensus earnings — is elevated by historical standards but more defensible than it appears when growth rates are factored in. The PEG Ratio (price-to-earnings divided by growth) for Nvidia remains at or below 1.0 on most consensus estimates, implying that the market is not obviously overpaying for the growth on offer.

The Long-Term Case: Infrastructure of the Intelligence Age

The BofA note, read carefully, is not simply a near-term trading call. It is a statement of conviction about the structural trajectory of AI as an economic phenomenon. A $1.7 trillion AI data centre market by 2030 is not a speculative fringe scenario — it reflects the observable capex commitments of the world's largest technology companies, the sovereign AI programmes being launched by governments from Saudi Arabia to Japan to the United States, and the increasingly visible productivity and revenue impacts that AI is generating across industries from financial services to healthcare to manufacturing. Nvidia, as the dominant supplier of the compute substrate on which this infrastructure is built, sits at the intersection of every major AI investment trend simultaneously. The BofA revision to $320 is, in this light, less a prediction about a stock price and more a statement about the scale of the opportunity that remains in front of the company.

Research Edition  |  For informational purposes only. Not investment advice. Always consult a qualified financial adviser before making investment decisions.