In this issue-

• Drone delivery of packages is here but regulation is still catching up

• Nuclear reactor miniaturization nears commercialization; 4 startups reached criticality in the last few weeks; could be positive news for AI data centers

• Satellite direct-to-device (phone) connectivity is approaching mainstream. It’s an emerging white space for startups

In each case, as you will soon read, the underlying innovation has been around for some time. But recently, a formerly slow moving external gate, whether a regulator, a standards body or a spectrum allocation, is opening up and years of technological progress seemingly rushing out.

So, each of these trends has reached an inflection point and is now going live and getting competitive.

Subscribe now

Drone Delivery is here; regulation isn’t

DoorDash was all over the news a couple of weeks ago for announcing that it was launching its drone delivery service, DoorDash Air developed by its in-house robotics unit, DoorDash Labs.

Of course this does not replace the cars, bicycles or rolling on-the-ground robots but it is one more option for customers in eligible neighborhoods in the coming months.

The FAA also regulates drone package delivery and DoorDash is the eighth drone operator to get certified for it by the FAA’s Part 135 Certification. Amazon Prime Air also holds this certification and has been in operation for some time and expanding fast.

What’s next

So the obvious ‘what’s next’ here is that more companies will consider drones as a last mile delivery method. Some will do it vertically, like Amazon, basically an extension of its own logistics network. Others like DoorDash will probably open it up in the future to other drone companies which can plug into its ecosystem.

That strategic choice, however, on how to bring drones into your business model, exists because of how hard that certification is to get. And that certification becomes the bottleneck as this method propagates for at least these reasons.

• The process is time-consuming and cumbersome. There are five phases to the FAA certification process. Some reports state that it takes 12-18 months and incurs considerable legal fees and paperwork to get through it. It covers maintenance programs, safety procedures, training and other aspects. That explains, at least partly, why there are very few companies that have gotten through so far.

• Part 135 certification wasn’t designed for drones. It was written for smaller aircraft, charter planes and cargo carriers. It’s now being retrofitted for drone delivery certification through exemptions and waivers. Unless the FAA creates something specifically for drone delivery, there is always going to be a bit of a misfit, which makes it even more complicated.

So whichever company can push through and get certified creates a moat of sorts. That means that there will be a few certified operators with drone makers supplying them with hardware rather than launching their own delivery systems.

Sure, that moat or lead could be temporary. The barriers to certification could eventually lower if the FAA adapts its existing framework. It has done it before when it accommodated a new aircraft category for powered lift for electric vertical takeoff and landing (eVTOL) air taxis like Joby not too long ago.

If that happens it will open it up to more delivery companies to get certified and run their own operations. In fact, like the eVTOL companies have done, early companies like DoorDash can contribute and help standardize what the FAA expects from a drone operator as they scale up.

But for now, the certification-driven moat stands.

Share

Nuclear Reactor Miniaturization Nears Commercial

Nuclear reactors generate electricity by splitting atoms (fission) to release heat. The heat turns water into steam that spins the turbines to produce power.

The way that nuclear power used to work is that you built one nuclear power plant at one site. This required several years of construction and of course millions of dollars. And since you poured cement in the ground for this one, it wasn’t portable.

In the last decade, a bunch of startups have been working to change that model. Or more accurately, miniaturize that model. They have been trying to mass produce small, affordable, and portable nuclear reactors. Companies like NuScale (2007) and Oklo (2013) have been working on this mostly on the R&D front for Small Modular Reactors (SMRs) for some years now.

Recently, there’s been a twist in this tale which has drastically sped things up.

In May 2025 the government created the Department of Energy’s reactor pilot program. This is a fast-tracked testing pathway that lets companies build and run test reactors by skipping the traditional, highly time consuming Nuclear Regulatory Commission licensing process. But, it came with a goal that at least three advanced reactor designs reach criticality by American Independence Day 2026.

Criticality means that it reaches a stage where a nuclear chain reaction is releasing enough neutrons to keep splitting atoms on its own, i.e. self sustaining. That’s the first physics milestone that every reactor needs to clear so that it can generate power.

New 1-3 year old startups like Antares Nuclear, Valar Atomics, Deployable Energy and Aalo Atomics used the fast-track lane and reached criticality in the past few weeks. By this point, all these startups are pretty well funded to continue to move forward.

What’s next

Reaching criticality is certainly an important goal but it mostly a minimum viability check. A proof of concept that proves that the physics works. But to take it to market is a different process.

To start producing electricity for customers, it needs to get licensed by the Nuclear Regulatory Commission (NRC). That is a still slower, multi-year process even if the NRC recently created its new ‘Part 53’ rule to streamline advanced reactor approvals. Meanwhile because of demand for energy, there is no dearth of customers right now.

There are two target markets for these products. The first is in the defense sector, where Antares and Deployable Energy are looking to make reactors that can be shipped on trucks or in containers to remote military bases and off-grid posts which need power on location.

The second market is created by energy demand at data centers. Valar Atomics is targeting this one. Its reactors are smaller but not portable. It clusters hundreds of them at fixed sites for data centers and other industrial customers, kind of the same way you’d stack servers rather than ship each one out. Valar has actually demonstrated this with its reactor powering an NVIDIA system and has a 30 MW AI data center deal locked down.

Although these are separate markets, my prediction is that the AI-powered one commercializes and scales faster than the defense one. Mostly because the demand is growing very fast. There is an energy shortage and the hyperscalers and other AI companies are going to any lengths to get the data centers up and operating quickly.

So interestingly enough, as nuclear reactors make the journey from cement-in-the-ground-plants to the portable devices model, the intermediate step of stackable-on-site model might be the initial winner.

Subscribe now

Satellite Direct-to-device (phone) connectivity approaches mainstream

The satellite-connected phone as a concept has been around for some time. These dedicated satellite phones were bulky and expensive. But they served their purpose and worked outside the range of a cell tower. So if you went hiking on some mountain with poor cellphone reception, the separate satellite phone device with its external antennas and customized internal components had you covered.

Now, a new and improved version of that, direct-to-device (D2D) satellite connectivity for our phones, eliminates the in-betweens. It doesn’t need cell phone towers or special hardware or a separate device because low earth orbit satellites are turning into space-based cell towers.

This has been made possible because two changes took place. First, 3GPP the body that develops the technical rules for mobile networks added satellite support called Non-Terrestrial Networks or NTN into the 5G standard. So most phone makers can use standardized chips instead of building custom chips for every different satellite provider (with some exceptions such as newer iPhones).

Second, satellite operators bought the specific radio frequencies that are required to communicate with our regular phones which made this service available more easily.

So now satellite direct-to-device connectivity is live.

As one example T-Mobile’s satellite service, which is powered by SpaceX’s Starlink, is available nationwide in the U.S. It offers texting, some limited app data and low-bandwidth voice and video on smartphones. However, these are not yet fully active for standard customers. Other providers like Verizon, AT&T plus carriers in Europe are getting ready to have their own commercial launches of these services.

Fresh off the press yesterday is the news below.

What’s next

For carriers, this is quickly becoming table stakes. As consumers, we can expect that sometime in the next year or two, when we go to change or upgrade our mobile phone service, there will likely be no premium add-on pricing for satellite connectivity.

It will ideally come bundled in (assuming this has been fully rolled out by then). The dead zones that were common with tower based cell service will start to disappear. That’s good from a safety standpoint but it’s also good for enterprise customers which rely on communications for remote logistics and other systems.

But the bigger shift, IMO, is that this is an inflection point because companies will start thinking about how to build on top of this infrastructure.

That’s what the Meta and AST SpaceMobile project focused on as shared in the posts above and below this section. It explored whether something like WhatsApp can be routed over satellite.

This also creates opportunities for startups to innovate and develop consumer apps or services where this becomes part of their ‘invisible’ technology stack.

Share Innovate & Invest

That’s all for this issue.

If you enjoyed reading it, please share with others in your network.

In this issue-

• New drugs, new logistics demands, and a delivery technology that could change it

• 3D printing could flip how batteries are designed; a white space for startups

• Edge computing is evolving into something smarter → Edge AI

Subscribe now

New drugs, new logistics demands

GLP-1s like Ozempic and Wegovy are changing lives and businesses in dramatic ways. As more people use them to tackle obesity, addiction, and a growing list of health conditions and prices come down, it is indirectly giving a boost to the logistics industry.

GLP-1 need to be stored and shipped in temperature-controlled facilities throughout the supply chain. Other new drugs coming down the pipeline are categorized as biologics. It means that they are derived from living matter and therefore temperature sensitive, requiring refrigeration from manufacture to delivery.

Demand for these drugs is expected to reach a market value of $39 billion by 2033. UPS has responded with a $48 million investment in 27 temperature-controlled facilities across the world.

On the other side is a countervailing force. Microneedle skin patches, currently commonly used for hormones and nicotine, are being developed as a way to deliver more and advanced drugs transdermally (i.e. through skin patches) This market is expected to double from current levels and reach $1.6 billion by 2036.

If the technology matures sufficiently, it will reduce the dependency on complex cold chain logistics. Its benefits include minimally invasive, pain-free self-administration, but there are some regulatory hurdles that companies making these have to overcome. For now, this technology is still in its infancy.

What’s next

These patterns are not new. Vaccines and insulin both forced the expansion and specialization of cold chain infrastructure in their time. Now it is happening again at a larger scale and with a twist of even more specialization with GLP-1s and biologics. Microneedle patches, when they come due, could potentially require yet another rethink about logistics.

Since these are both emerging trends, here’s what I think will happen.

The cold chain logistics boom that is being driven by GLP-1s and other new generation drugs is likely to sustain for some years.

However, while cold chain infrastructure will likely prevail for complex biologics and drugs that cannot be delivered any other way, drug delivery through micro-needle patches will gradually capture its own segment of the market for simpler self-administered treatments and mass vaccination…..but only after those companies have successfully navigated regulatory approval and scaled up manufacturing.

Share

3D Printing Redesigning Batteries

Today, batteries come in standard fixed forms. Any innovations in battery making in the last thirty years have been in the chemistry of the batteries, i.e., the ‘content’ and not as much the physical aspect. Products, whether drones, phones or or any other devices, have been designed around the battery shape and form.

Until now. Enter 3D printing.

According to a recent Wall Street Journal article, researchers published a whopping 25,000 papers in 2025 on the 3D printing of batteries!!!!!

The idea is that we can flip the way that products have been designed. Instead of building products and organizing components around batteries, 3D printing allows us to print batteries around the components, filling the available space between and around them in a product. 3D printing is ideal for the latter because it entails additive manufacturing, which builds an object by adding material layer by layer until you have the final shape.

The advantage is that 3D printing would enable energy storage to be integrated directly into products, using previously wasted space inside devices, drones, vehicles, and wearables; e.g. “Smartglasses could have sleek battery-packed frames”. So, increasing energy capacity without making products larger or heavier. It is agnostic to battery types which means that it can be used for lithium-ion, sodium-ion, solid-state and possibly other upcoming battery types.

Application of this technology is in its very early stages. There are barely a handful of startups that are innovating in this space. A couple of examples are Material Hybrid Manufacturing and Sakuu. That means there is plenty of room in the market for other start-ups to take this on.

What’s next

Since it’s commercialization, 3D printing has proven its value in high complexity, low volume, precision applications such as aerospace, medical devices, and hearing aids. It has significantly reduced the time and cost of product development across manufacturing sectors.

If it is able to achieve that again here, it would be another instance of manufacturing innovation unlocking design freedom….similar to how 3D printing has transformed industries before.

We can expect that, based on existing patterns for many new innovations, the first commercial applications will emerge in military and aerospace before eventually reaching consumer devices.

Share

Edge Computing to Edge AI

As someone who has worked with data for 20+ years, I can confirm that having more quantity and variety of data is usually a good thing. You can get more reliable results. That’s great when that data is being produced and processed on site. You can get fast results that humans can review and act on immediately, if that is your goal.

But if you are looking to process data received in real time from smartphones, autonomous vehicles, or industrial sensors that are located far away from where the data is processed, there is latency in receiving that data and then transferring results back to the source to take action. Bandwidth is usually limited and expensive.

The solution to this issue has been edge computing in which data is processed locally by following pre-programmed, fixed rules or algorithms. Locally means on the device or location itself. It reduces the bandwidth required to send all of it back and forth. Only relevant information and results are sent for further human review. It’s fast and efficient.

As a simple example, a factory sensor monitors temperature and is programmed to trigger an alert if it exceeds 200 degrees F. It doesn't send every temperature reading to the cloud or server. It just acts when the threshold is crossed. No human in the loop for regular temperature readings.

Now that we have developed advanced AI capabilities, edge computing is starting to morph into Edge AI. Here, the device isn’t just processing, it is also reasoning. It can handle ambiguous unstructured inputs like images or natural language and make judgment calls.

Edge AI is becoming a pattern across industries. See this fascinating use case of edge AI in space, literally the farthest point possible. In April 2026 for the first time ever a satellite in orbit used a vision-language model to identify “areas of interest in response to natural language queries.” It responded dynamically to something it wasn’t explicitly programmed to find.

What’s next

Well, this one writes itself. As more devices, vehicles, sensors, and yes, satellites generate data in places where transmitting everything back and forth to a central point is impractical or prohibitively expensive, intelligence (AI) will continue to shift to the edge.

The pattern is the same one we saw with edge computing but with AI capabilities added in. We can expect to see it show up in healthcare devices, agricultural sensors, and remote infrastructure monitoring, basically anywhere that data is produced far from where decisions need to be made. At some point as AI Agents become even more commonplace, edge AI could extend to acting on the results autonomously, without human review at all!

Share

If you enjoyed reading this article, please share with others in your network.