The Physical AI Pivot in Biotech Manufacturing

Fred Parietti stands in a warehouse in San Francisco’s Dogpatch neighborhood and says it plainly: someone has to make the drugs. This is not a vague metaphor about labor shortages. It is a hard engineering problem that has stalled the promise of gene therapies and mRNA treatments for a decade. Multiply Labs just closed a $75 million Series B round to solve it, and the money is not going into another algorithm. It is going into steel, test tubes, and robotic arms.

The funding brings Multiply’s total capital to over $100 million since its founding in 2016. The round was led by NantWorks, with participation from AstraZeneca, Lux Capital, and Founders Fund. This is a signal that the venture capital narrative is shifting. We spent the last few years pouring billions into AI models that can predict protein structures. Now, the smart money is betting on the machines that can actually bottle the result.

The Gap Between Discovery and Delivery

For years, the biotech sector has been defined by a strange disconnect. We have world-class computational tools for discovering molecules, but the production line remains stubbornly analog. Gene therapies and mRNA treatments are still largely made by hand in laboratories. The process is slow, difficult to scale, and dangerously vulnerable to human error and contamination. According to a 2025 research paper, manufacturing a single dose of some gene therapies can cost more than $1 million.

That price tag is not just a cost of goods sold issue. It is a barrier to access. Parietti argues that the only way to hit the necessary throughput, cost, and quality standards is with robotics. He wants to drop the price from millions of dollars to tens of thousands. To him, this is not a luxury upgrade. It is the difference between a drug that exists on a slide and a drug that saves lives at scale. The manual bottleneck is the final wall standing between scientific breakthrough and public health utility.

Robots in the Dogpatch Warehouse

Visit Multiply’s office and you do not see rows of sleek, minimalist workstations. You see a large, long box packed with test tubes and robotic arms. This is the physical reality of the company’s mission. Each robot costs several million dollars to build and relies on a supply chain spread across Europe. They are expensive, specialized, and currently deployed in a small number of facilities.

Parietti admits the company is still early. He describes the number of deployed machines as a high single-digit count. This is the critical risk in their pitch. They have proven the technology works, but they have not yet proven it scales. The next test is whether they can turn these bespoke, high-cost machines into a standardized manufacturing business that drugmakers can adopt without breaking their budgets. The gap between a prototype in a warehouse and a scalable industrial solution is where most hardware startups die.

The Rise of Physical AI

Multiply is not operating in a vacuum. It is at the forefront of a broader trend investors are calling physical AI. This is the category of AI systems that do not just process data in the cloud but act in the real world. The capital is flowing toward companies that can bridge the digital-physical divide. Multiply’s investors understand that the next wave of value in biotech will not come from a better model, but from a better factory.

The company’s business model is a hybrid. Drugmakers like AstraZeneca and Legend Biotech purchase the robots and operate them at their own facilities. Multiply also earns revenue from the cartridges used in the machines and from ongoing service and support. This creates a sticky relationship. Once a pharmaceutical company integrates Multiply’s hardware into its production line, the switching costs are high. The robots become part of the infrastructure, not just a tool.

Investor Conviction in Hard Tech

The composition of the Series B round tells its own story. NantWorks, the holding company founded by billionaire Patrick Soon-Shiong, leading the round is a statement of intent. Soon-Shiong is no stranger to long-horizon biotech bets. His involvement suggests a belief that the manufacturing bottleneck is a solvable engineering problem, not a fundamental biological limitation. Lux Capital and Founders Fund, both heavy investors in deep tech and AI, are also on the cap table.

This is a departure from the typical software-led venture capital cycle. These are investors who understand that hardware is hard, slow, and capital intensive. They are betting that the moat in biotech will be operational excellence. In an era where AI models can be replicated and open-sourced, a proprietary robotic manufacturing process is a defensible asset. Multiply is not just selling a product. It is selling a new standard for how medicine is made.

The Next Frontier

The stakes of this pivot are enormous. If Multiply succeeds, it does not just win a market share. It changes the economics of the entire life sciences industry. The cost of gene therapies and mRNA treatments could drop by an order of magnitude. That would unlock treatments for rare diseases that are currently unaffordable. It would make personalized medicine a reality rather than a theoretical ideal. The robots in the Dogpatch warehouse are the first step in that transformation.

But the path is not smooth. Supply chain constraints, the high initial cost of the machines, and the difficulty of scaling hardware all pose significant risks. Multiply has to prove that its robots can be manufactured as reliably as they can be deployed. It has to convince more pharmaceutical companies to take a chance on a new production paradigm. The $75 million is a vote of confidence, but it is not a guarantee. The real test is in the next five years, when the promise of AI-driven discovery must meet the harsh reality of industrial manufacturing.