ACCE AI Infrastructure Index sits at NAV 1287.43. Here's how NVDA, AVGO, META, GOOGL, MSFT, MRVL, and AMD stack up inside the index in mid-2026.
The ACCE AI Infrastructure Index sits at a NAV of 1287.43 as of July 10, 2026. That number tells a story: this index has compounded meaningfully since inception, and the composition explains why. Ten names. A deliberate mix of semiconductor muscle, hyperscaler capital deployment, and the networking layer that ties it all together.
This post breaks down what the index holds, how the weights are structured, and what the underlying data says about each position.
How the Index Is Built
The ACCE AI Infrastructure Index concentrates its top three positions equally: Broadcom (AVGO), Meta Platforms (META), and NVIDIA (NVDA) each carry a 15% weight. Alphabet (GOOGL) sits just behind at 14%, and Microsoft (MSFT) rounds out the top five at 12%.
The middle tier brings in Marvell Technology (MRVL) at 11% and Advanced Micro Devices (AMD) at 7%. The tail positions, Amazon (AMZN) at 4%, Cloudflare (NET) at 2%, and Datadog (DDOG) at 2%, add exposure to the infrastructure software and cloud networking layer without letting any single name dominate.
The logic is straightforward. The index is not a bet on AI applications. It is a bet on the physical and logical infrastructure that AI workloads require: chips, data center networking, hyperscaler compute, and the observability tools that keep it running.
The Anchor Positions
NVIDIA (NVDA) needs no introduction, but the 15% weight reflects a specific thesis: GPU compute remains the dominant training and inference substrate, and NVIDIA's software moat through CUDA compounds that hardware advantage. The stock carries an ACCE score of 63/100, with a Quality subscore of 96 and a Value subscore of 85. The 1-year return is -40.7%, which is the most important context here. NVIDIA at its current price of $221.35 is a very different risk/reward than NVIDIA at its 2025 peak. The 6-model fair value estimate sits at $442.09, implying +99.7% upside from the July 9 close. Revenue grew 12.7% year-over-year and earnings grew 7.9%. The FCF yield is 11.2% and net margin is 28.7%. The Momentum subscore of 17 reflects the drawdown, not the business.
Broadcom (AVGO) at 15% anchors the custom silicon and networking side of the index. AVGO's custom ASIC business has become the primary alternative to NVIDIA for hyperscalers building proprietary AI accelerators. Google's TPUs and Meta's MTIA chips both run through Broadcom's design and manufacturing relationships. No live price data is included in this post for AVBO beyond what the index provides, but the strategic rationale for the weight is clear.
Meta Platforms (META) at 15% is the index's most unusual inclusion for a casual observer. Meta is not a chip company. But it is one of the largest buyers of AI infrastructure on the planet, and its in-house AI work, from Llama to recommendation systems, makes it both a consumer and a driver of AI infrastructure demand. Its inclusion reflects the index's view that the hyperscalers funding the buildout deserve representation alongside the vendors supplying it.
The Second Tier: GOOGL, MSFT, MRVL
Alphabet (GOOGL) at 14% brings Google Cloud, TPU development, and DeepMind under one roof. Alphabet is simultaneously an infrastructure buyer, an infrastructure seller through Google Cloud, and an AI research lab. The 14% weight is the index's second-largest single-name bet.
Microsoft (MSFT) at 12% is the Azure story. The OpenAI partnership made Microsoft the default enterprise AI platform for a large portion of Fortune 500 companies. Azure's AI services revenue has been growing faster than the overall cloud segment. MSFT also appears in the ACCE Quality Compounders index at a 7% weight, which reflects its dual identity as both a compounder and an infrastructure play.
Marvell Technology (MRVL) at 11% is the index's highest-conviction pure semiconductor bet outside of NVIDIA. Marvell's custom silicon and optical interconnect business positions it directly in the data center networking upgrade cycle. As AI clusters scale from thousands to hundreds of thousands of GPUs, the networking layer becomes a bottleneck, and Marvell is one of a small number of companies that can address it.
AMD and the Tail
Advanced Micro Devices (AMD) at 7% is the index's hedge within the GPU space. AMD's MI300 series has gained meaningful traction with hyperscalers looking to diversify away from NVIDIA dependency. The 7% weight acknowledges the opportunity without overcommitting to an execution story that still has ground to cover.
The tail positions, Amazon (AMZN) at 4%, Cloudflare (NET) at 2%, and Datadog (DDOG) at 2%, serve a different function. They add exposure to the software and networking layer that sits on top of the hardware. Cloudflare's network sits at the edge of the internet and increasingly handles AI inference traffic. Datadog's observability platform is the default monitoring tool for cloud-native AI workloads. These are small positions, but they round out the index's coverage of the full infrastructure stack.
What the NAV Tells You
At 1287.43, the index has delivered real returns. The composition has not changed dramatically in recent months, which suggests the underlying thesis, that AI infrastructure spending is a multi-year capital cycle, not a one-quarter event, remains intact.
The NVIDIA drawdown is the most visible tension inside the index right now. A 15% weight in a stock that is down 40.7% over the past year has been a drag. But the quality metrics argue against reading that as a thesis break: 62.9% ROE, 28.7% net margin, 11.2% FCF yield, and a 6-model fair value that sits nearly double the current price. The market has repriced the stock on AI competition fears and export control uncertainty. The business has not deteriorated at the same rate.
The index's broader bet is that the companies building and enabling AI infrastructure, across chips, networking, cloud, and software, will capture a disproportionate share of the capital flowing into this cycle over the next three to five years. The NAV at 1287.43 reflects what has already happened. The composition reflects what the index expects to happen next.
Adobe (ADBE), which appears as a current ACCE pick with a Quality score of 96 and a 6-model fair value of $442.09 against a price of $221.35, is not in this index. Its infrastructure role is more ambiguous than its peers here. But the quality profile is worth noting for anyone tracking the broader technology landscape alongside this index.
The next major data point for this index comes from hyperscaler earnings, where capital expenditure guidance will either validate or challenge the infrastructure buildout thesis heading into the second half of 2026.
