Lenovo has introduced AI Express, an offering developed together with NVIDIA to speed up the rollout of artificial intelligence (AI) infrastructure in enterprises. The service combines validated server configurations, accelerated computing, and technical support to shorten deployment times. According to Lenovo, some configurations can go from order to shipment in 15 business days, while higher-capacity options start at 20 and 25 days, respectively.
Lenovo AI Express, the key points in 20 seconds
- The initiative offers three configurations for small, medium, and large-scale AI workloads.
- The small option starts at 15 business days; the medium one, 20; and the large one, 25.
- The systems pair Lenovo servers with NVIDIA GPUs, including the HGX B300 platform in the large configuration.
- Companies can add AI software, data protection, and services to take projects from testing to production.
Three configurations for different workloads
Lenovo AI Express relies on three quick-start configurations, sized according to model size, number of users, and required performance. The published lead times are commercial order-to-ship estimates for configurations that meet eligibility requirements, not a promise that the entire infrastructure will be installed and running within that time.
Small: starting at 15 business days
Inference for dozens of users
This configuration uses a Lenovo ThinkSystem SR650a V4 server with two NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. It’s aimed at models of between 7 billion and 70 billion parameters and claims performance of 30 or more tokens per second (TPS).
Medium: starting at 20 business days
Inference and agents for hundreds of users
The ThinkSystem SR675 V3 system includes eight NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. Lenovo positions it for models of between 70 billion and 400 billion parameters, with claimed performance of at least 30 TPS.
Large: starting at 25 business days
Deployments for thousands of users
The ThinkSystem SR680a V4 configuration is built on NVIDIA HGX B300 and is designed for generative models of up to one trillion parameters and large-scale enterprise workloads. Lenovo doesn’t set a single performance figure in the announcement for all scenarios in this category.
The three profiles let companies choose an initial capacity suited to their expected needs, rather than sizing a larger infrastructure from the outset. The idea is that businesses can expand resources as the number of agents, models used, users, and inference requests grow.
The TPS metric calls for some caution: actual performance depends on factors such as the model, input and output length, concurrency, and the service configuration. The 30 TPS figure quoted for the first two options shouldn’t be read as a universal guarantee for any model or workload.
Software, data protection, and support services
The offering isn’t limited to hardware. Lenovo lets customers expand the configurations with enterprise software, citing Red Hat AI Factory with NVIDIA and NVIDIA AI Enterprise among the options. It also includes Veeam Kasten to protect AI applications, data, models, and pipelines.
Combining these components aims to cover several needs of an AI factory: running models, integrating them into corporate processes, and protecting the data they use. The specific selection will depend on the architecture chosen and each company’s requirements.
Lenovo is adding consulting and deployment services to help identify use cases, evaluate return on investment through proof-of-concept testing, and tune GPU environments to improve utilization and control costs. It also announces new proactive and predictive server support capabilities through Premier Support Plus for Servers, aimed at catching problems earlier and keeping critical operations running.
The company frames this offering within its Hybrid AI Factory with NVIDIA concept, which aims to combine infrastructure, software, and services to run AI in enterprise environments. Deployments can address inference, generative applications, and agentic systems, though the right configuration will depend on the workload and on performance and security requirements.
Lenovo also wants to extend AI Express’s distribution through its international network of business partners and the Lenovo 360 program. Availability, eligibility criteria, and ordering procedures may vary by region, so the published lead times should be confirmed in each market.
The announcement cites Lenovo’s CIO Playbook 2026: according to the company, organizations that move from experimenting with AI to using it at scale expect to see an average return of $2.79 for every dollar invested, and 93% of large-enterprise respondents anticipate positive returns. These are expectations drawn from the study cited by the manufacturer, not guaranteed outcomes for AI Express customers.
Frequently asked questions
What is Lenovo AI Express?
It’s an offering from Lenovo and NVIDIA that bundles validated AI infrastructure configurations, optional software, and services to speed up moving AI projects into production.
How long does a Lenovo AI Express system take to arrive?
Lenovo quotes order-to-ship times starting at 15 business days for the Small configuration, 20 days for Medium, and 25 days for Large. These apply to eligible configurations and can vary by region.
What GPUs do the configurations use?
The Small and Medium options use two and eight NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, respectively. The Large configuration is built on NVIDIA HGX B300.
Can the infrastructure be expanded later?
Yes. Lenovo’s approach is to start with a configuration sized to initial needs and expand capacity as models, users, and inference demand grow.
