Three announcements landed within 48 hours of each other in mid-August, and together they show where enterprise AI budgets are actually flowing this quarter: dedicated inference capacity, long-term cloud commitments from large non-tech buyers, and cheaper, faster models built to run agent workloads at scale.
IBM and Together AI commit $240 million to inference capacity
IBM and Together AI announced a multi-year, $240 million agreement on August 11 to build a dedicated cluster of NVIDIA HGX B300 systems on IBM Cloud, according to IBM's newsroom announcement. The cluster will use NVIDIA Spectrum-X Ethernet networking and is the first large-scale deployment built specifically for inference on IBM Cloud using B300 hardware.
Together AI will run the cluster to serve open-source model inference through its AI Native Cloud platform, extending the service to enterprise customers who want open-weight models without managing their own GPU fleets. IBM said the deployment is expected to be available starting in the first quarter of 2027.
The deal is notable less for the dollar figure than for what it signals about IBM's positioning. IBM has spent the past two years talking about AI mostly through watsonx and consulting engagements. A capital commitment to raw inference infrastructure, in partnership with a company built entirely around serving other people's models, is a different kind of move: it puts IBM Cloud in direct competition for inference workloads against AWS, Azure and Google Cloud rather than positioning it as a governance layer sitting above them.
Ryanair locks in a five-year AI and cloud partnership with Google
Ryanair and Google Cloud announced a five-year data and AI partnership on August 12, according to a joint release published through Google Cloud's press corner and confirmed on Ryanair's corporate site. The airline will roll out Google Workspace and Google Cloud infrastructure to 35,000 employees and adopt Gemini Enterprise, Google's agentic AI platform, alongside Google DeepMind models.
Ryanair said the deployment is intended to support crew logistics, maintenance scheduling and broader operational decision-making as the airline scales toward a stated target of 300 million passengers by 2034. The agreement also gives Ryanair a second major cloud provider alongside its existing infrastructure, a dual-cloud approach the airline described as adding resilience.
This is a useful data point for anyone tracking how agentic AI platforms move from pilot projects into operational commitments at large, non-tech employers. Ryanair is not a software company. It runs thin margins on a business built around aircraft utilization and turnaround time, and a five-year AI and cloud contract at that scale is a bet that Gemini Enterprise can be embedded into scheduling and logistics decisions that directly affect the airline's cost base, not just its back office.
Google ships Gemini 3.7 Flash, priced for high-volume agent use
Google introduced Gemini 3.7 Flash on August 13, describing it in its own blog post as its most capable "workhorse" model for coding and agent workloads to date. Google reported gains on internal and third-party benchmarks, including an improvement on the DeepSWE v1.1 benchmark from 49.0 percent to 65.3 percent compared with the prior Flash model.
The pricing is the part worth watching. Google is offering Gemini 3.7 Flash at an introductory rate of $0.75 per million input tokens and $3.75 per million output tokens through the end of 2026, half the launch price of its predecessor. Standard pricing of $1.50 and $7.50 takes effect afterward. The model is available now through the Gemini API, Google AI Studio, Android Studio, Google Antigravity and the Gemini Enterprise Agent Platform, the same platform Ryanair is adopting.
The timing lines up with the infrastructure announcements above. Cheap, fast models are what make agent deployments at the scale Ryanair and Together AI's enterprise customers are describing economically viable. A model that costs half as much per token to run changes the math on whether an agent workflow that calls a model dozens of times per task is worth deploying at all.
What to watch next
Together AI's IBM cluster is not expected online until Q1 2027, which leaves several months for competitors to respond with their own dedicated inference capacity announcements. Google's introductory Gemini 3.7 Flash pricing is explicitly set to expire on December 31, 2026, a hard deadline that will show whether adoption holds once the discount disappears. And Ryanair's rollout, still in its early stages, is the kind of large-scale, non-tech enterprise deployment that will be worth checking on for concrete usage results rather than announcement-stage claims.
Sources
- IBM Newsroom, "IBM and Together AI Sign Multi-Year Agreement to Scale Open-Source AI Inference with NVIDIA AI Infrastructure on IBM Cloud" - https://newsroom.ibm.com/2026-08-11-IBM-and-Together-AI-Sign-Multi-Year-Agreement-to-Scale-Open-Source-AI-Inference-with-NVIDIA-AI-Infrastructure-on-IBM-Cloud
- The Register, "Together AI embraces the competition with $240m IBM Cloud deal" - https://www.theregister.com/off-prem/2026/08/11/together-ai-embraces-the-competition-with-240m-ibm-cloud-deal/5286400
- DataCenterDynamics, "IBM and Together AI sign $240m NVIDIA HGX B300 cloud deployment deal" - https://www.datacenterdynamics.com/en/news/ibm-and-together-ai-sign-240m-nvidia-hgx-b300-cloud-deployment-deal/
- Google Cloud Press Corner, "Ryanair and Google Cloud Announce Five-Year Data and AI Partnership" - https://www.googlecloudpresscorner.com/2026-08-12-Ryanair-and-Google-Cloud-Announce-Five-Year-Data-and-AI-Partnership
- Ryanair Corporate, "Ryanair, Google Cloud Announce Five-Year Data and AI Partnership" - https://corporate.ryanair.com/news/ryanair-google-cloud-announce-five-year-data-and-ai-partnership/
- Google Blog, "Introducing Gemini 3.7 Flash" - https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-3-7-flash/
- 9to5Google, "Gemini 3.7 Flash launch" - https://9to5google.com/2026/08/13/gemini-3-7-flash-launch/