CONFIRMED: Equinix, Nvidia and Together AI launch inference exchange for 2027

Equinix announced on September 2, 2026, a partnership with Nvidia and Together AI to launch Equinix Inference Exchange, a distributed AI inference platform aimed at enterprises. The announcement appeared on Equinix's official newsroom and was picked up by CNBC, Data Center Dynamics and other infrastructure focused outlets. The company made clear the service is not live yet: availability is expected to begin in the first quarter of 2027, which makes this an announcement of product intent, not an immediate launch.

What Equinix Inference Exchange actually is

In practice, Equinix Inference Exchange combines three pieces. First, Nvidia's Enterprise Reference Architectures, validated hardware and software blueprints for running inference workloads at scale. Second, Together AI's inference platform, which already supports more than 200 open source models and handles execution orchestration for them. Third, Equinix's physical footprint, which spans more than 280 data centers across 77 metro areas worldwide, connected through more than 230 direct links to cloud providers via Equinix Fabric. The pitch is letting enterprises run inference closer to where their data and users sit, cutting latency, gaining flexibility to move from closed to open models, and meeting data sovereignty requirements in specific countries.

Separately in the same week: Huang talks nearly $1 trillion in the US alone

Separately, but in the same window, Nvidia CEO Jensen Huang said, between September 3 and 4, 2026, at a G20 innovation ministers meeting in North Carolina, that Nvidia plans to invest close to $1 trillion in AI infrastructure in the United States alone this year. According to Seoul Economic Daily, Huang tied that figure to deregulation and pro energy policies in the country, and went as far as comparing AI's importance to that of water, roads and electricity. It is essential to treat that number for what it is: a target announced by the executive himself, not a figure audited by a third party or confirmed in a company financial filing.

Why it matters: capacity and price, not a new product today

For Brazilian small and medium businesses and for customer service and marketing automation professionals, the direct message is that none of this is a product to use tomorrow. Equinix Inference Exchange only arrives in the first quarter of 2027, and Huang's trillion dollar figure is investment in physical capacity (data centers, chips, power), not a new AI tool available today. What these announcements signal is a medium term trend: historically, more installed inference capacity tends to push down cost per token and API latency for the AI tools agencies already use daily, as infrastructure supply matures and competition among cloud providers, chipmakers and inference platforms increases. That is not a guarantee of lower prices in 2026 or 2027, it is a likely direction, not an immediate effect.

What to watch over the coming months

Three things are worth watching over the coming months: whether the arrival of Equinix Inference Exchange actually lowers the cost of accessing open models hosted through Together AI once the service goes live; whether other cloud and inference providers announce similar capacity commitments, which would reinforce the growing supply thesis; and whether Huang's $1 trillion figure shows up, in detail, in financial reports or specific project announcements from Nvidia, turning a pledge into a trackable commitment.

Direct take

The two stories from the same week tell one narrative from different angles: Nvidia is betting heavily on being the infrastructure behind all the AI that comes next, not just the chip supplier for training models but also the name behind where and how those models run in production, through partners like Equinix and Together AI. Equinix, in turn, is using the partnership to reposition itself as central to the inference era, after years of being sold mainly as a neutral data center provider. It is worth watching, with healthy skepticism, the actual pace of execution: between a partnership announcement and a trillion dollars actually spent lies a distance measured in years, not weeks.

Sources

Equinix Newsroom, Equinix Accelerates AI Inference for Enterprises with NVIDIA and Together AI: https://newsroom.equinix.com/2026-09-02-Equinix-Accelerates-AI-Inference-for-Enterprises-with-NVIDIA-and-Together-AI | CNBC, Equinix partners with Nvidia, carves niche in AI data center boom: https://www.cnbc.com/2026/09/02/equinix-partners-with-nvidia-carves-niche-in-ai-data-center-boom.html | Data Center Dynamics, Equinix, Together AI, Nvidia partner on Inference Exchange: https://www.datacenterdynamics.com/en/news/equinix-together-ai-nvidia-partner-on-inference-exchange/ | Seoul Economic Daily, Nvidia to Invest $1 Trillion in U.S. AI Infrastructure This Year: https://en.sedaily.com/international/2026/09/03/nvidia-to-invest-1-trillion-in-us-ai-infrastructure-this | Seoul Economic Daily, Nvidia CEO Pledges $1 Trillion for U.S. AI Infrastructure: https://en.sedaily.com/international/2026/09/04/nvidia-ceo-pledges-1-trillion-for-us-ai-infrastructure