Manufacturing the future: How Sanmina is scaling AI

In the world of high-tech manufacturing, complexity isn't just a challenge—it’s the baseline. For Sanmina, a Fortune 500 leader in electronics manufacturing, that complexity involves managing 60 global factories and hundreds of unique supply chains for everything from MRI machines to 5G base stations.

At Google Cloud Next, Manesh Patel, Chief Information Officer (CIO), Samina presented with Paul Lewis, Chief Technology Officer, Pythian discussed how Samina, a $10 billion manufacturing giant is moving past the AI hype to deliver real operational value. Manesh Patel describes Sanmina’s AI journey as a balanced Rubik’s Cube of innovation. They aren't just chasing one silver bullet solution; they are implementing AI across four distinct functional points.

Who is Sanmina?

While you might not see the Sanmina name on your household appliances, they are the "engine room" for some of the world’s most recognizable OEMs. They specialize in mission-critical electronics across:

• Medical: Complex systems like MRI and ultrasound machines.
• Communications: 4G and 5G base stations.
• Industrial & Energy: Clean energy solutions and industrial tech.
• Defense & Aerospace: High-reliability components.
• Hyperscale Data Centers: Through their acquisition of ZT Systems, Sanmina now manufactures the massive, high-density server racks that power the cloud (including Google’s own infrastructure).

Pillars of AI innovation at Sanmina

Personal productivity

Sanmina has deployed Gemini for Google Workspace to over 23,000 knowledge workers. By automating routine tasks—drafting emails, summarizing meetings, and organizing data—they have already recaptured an estimated 350 hours per week in productivity.

The Pro Tip: To speed up the development lifecycle, Sanmina teams now voice-code their requirements. They transcribe a meeting in Google Meet and use Gemini to generate the first version of a technical requirements document instantly.

Mastering the supply chain with Vertex AI

With over 100 customers each having their own supply chain, the math is staggering. Sanmina is piloting Google Vertex AI for supply-demand matching. The goal is to analyze component availability and factory capacity simultaneously, allowing them to give customers a firm commitment on a forecast in just 2-3 days—a calculation that used to take much longer.

Operational technology (OT) and data quality

In manufacturing, the OT (the sensors and equipment on the factory floor) creates 20 times more data than standard IT systems. Sanmina’s advantage lies in its standardized ERP and manufacturing systems across all 70 sites. This high-quality, clean data is fed into BigQuery, providing a single source of truth that makes AI models more accurate and easier to deploy.

Manufacturing for the AI era

Sanmina isn’t just using AI; they are building the hardware that makes it possible. As GPUs become more powerful, server racks are jumping from 100 kilowatts to a projected 300 kilowatts in the next year. Sanmina is at the forefront of designing the liquid cooling and power distribution systems required to keep these AI brains from overheating.

Governance: Federated innovation, centralized guardrails

How do you keep 39,000 employees from creating a wild west of unsecure AI? Patel’s approach is federated innovation:

• AI office: A centralized group at the CIO level that sets policy and security standards.
• Controlled access: Access to external AI tools is locked down by default, funneling users toward the secure Google Gemini environment.
• Incentivized adoption: Rather than forcing use, the IT team identifies laggards and provides targeted training to show how AI can solve their specific pain points.

Manesh Patel closed the talk with a reality check for other leaders: AI is harder than the media makes it look. His two pieces of advice for organizations starting their journey are:

1. Educate the executive suite: Ensure leadership understands the how and why of AI, or approvals will stall.
2. Start small, then scale: Don't rush to put autonomous agents everywhere. Start with small wins, climb the learning curve, and build a foundation of high-quality data first.

By focusing on the design time rather than just the runtime, Sanmina is proving that the biggest wins in AI often come from making the complex work of human collaboration just a little bit simpler.

Fast-track months of strategic conversations—clearly define your data and AI strategy.