Browse > Home / AI, Business process, Business strategy, News / AI: Is the Next Big Thing for Manufacturing? But Whose?

AI: Is the Next Big Thing for Manufacturing? But Whose?

March 18, 2026

green-dino_thumbAnother dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.

For the second time in three days, I have heard about the “physical problem of AI.” Each write up defines “physical” implicitly. Both, however, view AI in the context of a mental view like this: Tomorrow will be like yesterday.  What does this mean? The implication is that AI software is ready to go big time but for:

  1. Power. Yes, it takes time to build power generation facilities. Even getting a gas turbine and hooking it up to an existing data center can take months. How long does it take to build a baby nuclear plant? I don’t know, but Bill Gates’ venture will answer that question? Why not just buy the reactors that power nuclear vessels like submarine? Sorry, that’s classified.
  2. Chips. Yes, that’s a bit of a problem  with at least two dimensions. The first is that only a few companies manufacture the systems that pump out semiconductors. Supply chains exist for these vendors, and some of the gizmos required by ASML-type outfits cannot be purchased from Amazon. The second dimension is the pace of change of the chips themselves. Each time an Nvidia-type company rolls out a new AI chip, the data centers purpose built for previous generations of AI chips need tweaking. What happens when many people want to tweak at the same time? Cost spikes and bottlenecks, perhaps?
  3. Data centers. Telegram, piloted by the clever and likeable Brothers Durov, concluded that using other people’s infrastructure or OPI was a better way to do AI. No capital investment, no verticalization, and no physical facilities on one’s books — That is the way to go. The direction is the opposite of some other big tech outfits. Are the Brother Durovs correct? some bean counters think the duo may be.

I thought about these factors when I read the marketing collateral published as “real” information in MIT Technology Review. (No, I won’t mention Epstein Epstein Epstein again in the context of the estimable institution.) The article is titled “Why Physical AI Is Becoming Manufacturing’s Next Advantage.” Okay, tomorrow will be like yesterday because new technology revolutionizes something fundamental like manufacturing.

The idea is unrelated to the three points mentioned above about the problem physical infrastructure poses for big tech AI. Hey, folks, where are you going to get helium? What will be the cost of components when disruption hammers companies in those supply chains for the semiconductor and infrastructure companies? Industrial machines have supply chains too, and these may not operate as they did yesterday.

image

Yesterday many not prepare some for today. Today may not predict tomorrow. Foundational assumptions require identification, analysis, and fact-based predictions. Thanks, Venice.ai. Good enough for handling medical treatments and the manufacturing of experimental modular nuclear reactors for the data centers to be constructed in this open field in Poland.

The write up asserts:

The next phase of transformation will not be defined by isolated AI tools or individual robots, but by intelligence that can operate reliably in the physical world. This is where physical AI—intelligence that can sense, reason, and act in the real world—marks a decisive shift. And it is why Microsoft and NVIDIA are working together to help manufacturers move from experimentation to production at industrial scale.

I like the notion of “industrial scale.” It is a harbinger of big money for those who capture the market. What’s going to create this conquering of the “industrial frontier”? Answer: Microsoft and Nvidia. Microsoft, the MIT Technology Review helpfully points out, is the author of the article. The argument flows in a PR-ish rush that “physical AI” will arrive at scale, enabled by Microsoft and Nvidia. Instead of talking about this revolution, these two firms will move from intelligence to action.

As I have pointed out, humans want to believe that today’s AI is indeed good enough. The idea has fueled adoption of smart software by a number of firms. Even governments have found value in today’s smart software. As some forward leaning people assert, tomorrow will be like yesterday. I would suggest that:

  1. The coming supply chain disruptions will make such assumptions subject to endless revision and revisionism
  2. The friction of the humans involved in these predicted inevitable shifts may add both time and emotional heat to the inevitable change
  3. The big outfits confidently predicting the future from their vantage point may be operating on false assumptions.

I noted this passage in the advertisement for the Microsoft and Nvidia view of the industrial future:

As physical AI systems scale, trust becomes the limiting factor. Manufacturers must ensure that AI systems are secure, observable, and operating within policy, especially when they influence safety?critical or mission?critical processes. Governance cannot be an afterthought; It must be engineered into the platform itself. This is why frontier manufacturers treat trust as a first?class requirement, pairing innovation with visibility, compliance, and accountability. Only then can physical AI move from promising demonstrations to enterprise?wide deployment.

Infrastructure, supply chain issues, and geopolitical instability — Not a problem. The problem is trust; it is the limiting factor. Does Microsoft engender trust? I am wary of quasi-monopolies engendering trust in dinobabies like me. I don’t “trust” anyone or any thing unless either unless I have verified it in accordance with my old-fashioned principles. QR codes on my mobile for a menu? Nope. Cough up information to use a free service? Nope. Believe that Microsoft can update its software without crashing an essential function like printing? Nope. Believe that Nvidia’s endless stream of driver updates improve my experience? Nope. Your mileage may vary, or you may rationalize by saying to yourself, “That’s the way it is.”

And security? I am not going to recycle my comments, findings, and perceptions of Microsoft-type companies’ security. Security increases these firms’ costs. What these firms require is margins, profits, and bonus pools spilling cash into Carpertland.

Will manufacturing (whatever that means to Microsoft’s PR team) embrace AI? The answer is, “Over time.” Will manufacturing in China use AI? Answer: Many Chinese firms are deeply integrating AI into certain processes. Isn’t that why Apple loves to do business with certain Chinese manufacturers or is it because those Apple executives love driving in Beijing, Shanghai, or Suzhou for meetings?

Here are my customary observations about this PR piece designed to make General Motors-type outfits immediately plan to acquire machines to replace those silly manual gizmos in the model shop and the endangered tool shops in the firm’s plants and spark the purchase of new smart machines for the planned Sunday Mines Complex. Machines are less expensive than the legal fees for worker health claims, a troubling issue for bean counters. AI to the rescue.

  1. Timing. Microsoft’s claims about AI and manufacturing are the PR equivalent of a farmer tilling before planting. The fallacy is that tomorrow will be like yesterday.
  2. Cost. An issue now and tomorrow because the geopolitical instability does not decrease economic uncertainty and make bankers think about distant time horizons or risk averse investors just dive in as they did yesterday.
  3. Reliability. The outputs of probabilistic word prediction-centric systems will produce errors some percentage of the time. Is good enough going to be good enough? That’s an interesting question because the answer depends on context.  How about those smart systems prescribing chemicals for your daughter’s chemo treatment? Are you onboard?
  4. Stable supply chains.  Yesterday, maybe? Today, perhaps if you pay up front and buy what’s available? Tomorrow? Why not predict the winner of the 2026 Kentucky Derby? If you leave soon, you can visit the Hormuz area and maybe return before the race.

Net net: There are now quite significant restraints on the current smart software sector. Some are pragmatic like power and water and gizmos. Others are downstream like industrial machines with AI inside. The flaw is that tomorrow is starting to look less and less like yesterday.

What country will do the AI in manufacturing thing? Polymarket?

Stephen E Arnold, March 18, 2026

Written by Stephen E. Arnold · Filed Under AI, Business process, Business strategy, News 

Comments

Got something to say?





  • Archives

  • Recent Posts

  • Meta