NVIDIA Secures 37% of 2027 HBM Output Amid Record AI Infrastructure Buildout

Strategic block trades and unprecedented capital commitments signal institutional scramble for advanced AI capacity and power, reshaping semiconductor and energy markets.

Tradesnaut Quant Research Desk · September 28, 2026 · 6 min read · AI Market Analysis

NVIDIA Secures 37% of 2027 HBM Output Amid Record AI Infrastructure Buildout

Key takeaways

Leading semiconductor companies are locking in future high-bandwidth memory (HBM) supply through multi-year commitments, creating a scarcity for uncontracted buyers.
The surge in AI infrastructure spending is driving unprecedented M&A activity in the power and utilities sector as hyperscalers seek to secure dedicated energy sources.
Watch for further consolidation in both semiconductor memory and energy generation, alongside potential shifts in AI demand forecasts or HBM manufacturing advancements, which could alter capital flow patterns.

What changed

A significant shift in institutional capital allocation is evident in the semiconductor and energy sectors, driven by the escalating demands of artificial intelligence. NVIDIA, a bellwether for AI computing, has strategically locked in commitments covering approximately 37% of the global high-bandwidth memory (HBM) output for 2027, according to supply-chain procurement data compiled on September 25, 2026. This substantial allocation, along with commitments from Alphabet and Advanced Micro Devices, accounts for about 85% of global 2027 HBM production capacity, as estimated by Morgan Stanley Research. For uncontracted buyers, this means facing higher component prices and extended lead times through at least 2027.

Broadcom has also solidified its position in the custom AI chip market. The company reported fiscal Q3 2026 revenue of 9.6 billion, an 86% year-over-year increase, with AI revenue surging 221% to

6.7 billion. While Q4 FY26 revenue guidance of

4.8 billion was slightly below some LSEG analyst consensus estimates, Broadcom raised its full-year AI revenue forecast for fiscal 2026 to $58 billion from $56 billion and projected AI semiconductor revenue of

15 billion for fiscal 2027 and 30 billion for fiscal 2028, having secured the necessary supply. This strategic positioning, often facilitated by off-exchange block trades that reflect deep institutional conviction, underscores a profound industrial transformation. Meanwhile, the energy sector has witnessed record M&A activity, with power and utilities mergers and acquisitions reaching 05 billion in the first half of 2026, more than tripling the value from the same period in 2025.

The mechanism

The central mechanism at play is the critical scarcity of high-bandwidth memory, which is essential for powering modern AI accelerators. HBM chips, with their vertically stacked memory dies, provide the immense data throughput required by large language models and other AI workloads. Manufacturing these advanced components is complex and capacity-constrained, leading major players to secure supply years in advance through significant capital commitments and pre-payments. NVIDIA's Q2 FY27 regulatory filing on August 27, for instance, disclosed a surge in supply and capacity commitments to 79 billion, up from

19 billion in the preceding quarter, reflecting these advance contractual obligations.

This demand for AI compute extends directly into the energy sector. Running massive AI data centers requires prodigious amounts of power, straining existing grid capacities. As a result, AI-driven load growth has become a central driver of M&A activity in power and utilities. Hyperscalers are increasingly looking to invest directly in generation assets to ensure a stable and sufficient power supply for their AI factories. This confluence of technology and power sectors is creating a new M&A cycle, evident in landmark deals such as NextEra Energy's $67 billion acquisition of Dominion Energy, an acquisition valued at $420 billion enterprise-wide. Valuations for gas-fired generation M&A, for instance, have more than doubled from $675 per kilowatt in 2024 to

,468 per kilowatt in the first half of 2026, demonstrating a scarcity premium for dispatchable power.

Who is exposed

The companies most exposed to these trends are the key players in the semiconductor and energy infrastructure spaces. NVIDIA (NVDA) and Broadcom (AVGO) are clear beneficiaries, with their long-term supply agreements and strong positioning in AI hardware. NVIDIA reported a record Q4 FY26 revenue of $68.1 billion, a 73% year-over-year increase, primarily driven by its Data Center segment, which saw a 75% year-over-year rise to $62.3 billion. AMD (AMD), while also expanding its AI presence with products like the Instinct MI300X, which boasts 192GB of HBM3 memory, faces intense competition for HBM supply, especially against NVIDIA's aggressive lock-ins. AMD's Q2 2026 revenue grew 50% year-over-year to

1.54 billion, with data center sales up 107% year-over-year. The company expects Q3 2026 revenue of

3 billion.

In the energy sector, utilities and power generation companies with ample or expandable capacity are positioned for significant M&A activity and direct investment from tech giants. The current market dynamics favor those who can secure consistent and reliable power. Conversely, companies attempting to build proprietary AI hardware or expand data center capacity without established multi-year HBM supply agreements or secure energy partnerships face considerable procurement risks and cost disadvantages. The competitive landscape for smaller or new entrants in the AI chip and infrastructure space is tightening dramatically.

Quantitative Outlook

The market data for semiconductor equities reflects the intense investor focus on the AI theme. NVIDIA (NVDA) closed at 229.20, up 1.83% for the day, and has seen impressive year-over-year growth of 29.29%. Broadcom (AVGO) closed at 348.85, down 1.12% for the day, reflecting its Q4 guidance falling slightly below some consensus estimates, yet its 1-year performance shows a 4.57% gain. Advanced Micro Devices (AMD), despite a daily drop of 5.22% to 597.74, boasts a remarkable 270.65% increase over the past year, indicating strong long-term conviction in its AI story (see the table below).

These individual stock movements are occurring within a broader market context where the S&P 500 is down 0.94% to 7,670.97 and the NASDAQ is down 1.19% to 26,747. The relative strength in these semiconductor giants, particularly over longer timeframes, suggests that institutional capital remains firmly committed to the AI growth narrative, even as the broader indices experience downward pressure. The off-exchange order flow, while difficult to precisely quantify without proprietary dark pool data, is indicative of large, strategic positions being built quietly. The sustained demand for AI infrastructure, projected by Gartner to drive worldwide AI-optimized IaaS spending up 96% to $42 billion in 2026, reinforces the long-term runway for these companies. Further clarity on HBM4 production yields, new entrants in the AI chip market, or shifts in hyperscaler capital expenditure priorities would be crucial to monitor. However, the current picture points to a sustained, concentrated investment cycle into the foundational components of the AI revolution.

Tags: NVIDIA, Broadcom, AMD, Semiconductors, Energy, AI Infrastructure, Block Trades