Boreal Tech Brief #052

SPONSORED · NFA

NFA just raised SEK 48 million.

I covered the raise in Tuesday's brief, and I sat down with NFA to go deeper. That conversation led to this sponsored deep dive. The story stands on its own: what the money is for, and where they want to push the technology next.

It is marked as sponsored throughout, and you will see where my own read comes in.

Axel

A thinning harvester operator makes more decisions per minute than a fighter pilot. The comparison sounds like a slogan, but Skogforsk, Sweden's forestry research institute, has studied it. The operator reads each stem, weighs it against the goal for the stand, decides which trees come out and which stay, and then does it again, second after second. NFA's software is built to sit inside that loop. It records the stand as the machine moves through it, recommends which stems to take out, and then measures what is left standing. That is decision support, working alongside the operator on the call itself.

NFA, Nordic Forestry Automation, sits in Lund and builds both halves of the product. The software reads the standing trees as the machine works: location, size, form, and species. The hardware is theirs too, the sensors paired with the compute unit that runs the calculations, assembled by a local partner in Lund. The first product, live today, is operator support for thinning. It runs in commercial operations across seven markets on three continents, on roughly a dozen harvester models, with more than 20,000 operational hours behind it. Customers include Sveaskog, Södra and Holmen in Sweden, and Weyerhaeuser, the largest private landowner in the United States.

NFA frames its own road as three generations. Generation 1, today, is decision support: the operator stays fully in control while the system reads the stand and flags the trees to take. Generation 2, the one this round funds, is semi-automation. The machine handles part of the work on its own while a driver supervises and takes over the tricky parts. The closest model is the self-steering tractor, which has guided itself across fields in agriculture for a long time: a person still in the seat for the turns and the awkward moments, watching the rest. Generation 3, full automation, sits further out, and I will come back to it.

The SEK 48 million round, led by Navigare Ventures and Almi Invest Greentech, does two things at once. It pushes Generation 1 wider, into more machines across the markets NFA already serves, and it funds the build of Generation 2. The round also adds two owners, Almi Invest Greentech and LRF Ventures, alongside existing backers including Navigare, owned by Wallenberg Investments, Sveaskog and Södra.

The plan is not to plant flags on new continents. NFA already operates in Sweden, the Baltics, the U.S., New Zealand, Australia and Canada, and the stated focus is to go deeper in those markets. The point is to build on the ground they already hold rather than chase another continent. Assembly of the hardware stays in Lund, even as it might have to scale up hard.

Today the software sees the stand and recommends which tree to take. Getting the machine to act on that recommendation by itself is the Generation 2 problem. The main challenge NFA describes is control and integration: once the algorithm has picked the stem, how do you steer the crane and head to take that one tree? There is no shared industry interface for this, so every integration is specific to the machine brand, and NFA works model by model.

Under that sits a more basic limit. Asked what is hardest about the product, NFA said it is getting a correct picture of reality. I pushed on whether that is a LiDAR limit or a processing one. The answer was that it is the trade-off between them: more data points give a more detailed picture, but cost time to process, and the skill is finding the balance.

One Generation 2 capability is identifying tree species from LiDAR. In NFA's earlier work with a Swedish customer, getting species distribution right was a clear win. I asked how accurate it has to be. They are bound by customer agreements and could not name the exact target. When I asked whether it sits around 95 percent, they said yes. They are working actively to get there, and they sound confident they will.

The system also captures detailed data on the trees left standing after a thinning: location, size, form, species. NFA sees that data as a business in its own right. My read is that they own it and could sell it. What you could build on a dataset like that is the open question, and here the speculation is mine, not NFA's. Growth and yield models, the classical kind or the machine-learned kind, trained on a volume of real stand data that is hard to gather any other way. Forestry has long run on sample plots and statistical curves. This much ground truth could rebuild parts of it from scratch.

That value is not evenly spread, and NFA confirmed the pattern when I put it to them. In Sweden, where we already sit on dense, nationwide forest data and high-quality harvester data, the decision support carries more of the weight. In other markets, where that baseline is thinner, the balance tilts toward what the machines collect as they work. Same product, weighted differently by continent.

This next part is me thinking aloud, not NFA.

It is easy to picture an autonomous thinning machine as today's machine with an empty seat. Once you take the operator out, though, you also take out the reason the machine looks the way it does. The cost per hour of putting a person in the cab shapes the whole business model of a forestry contractor, and that cost is a big part of why machines have always been built around a driver. Take the driver away and you no longer have to design around the cab, or around the envelope built to keep one person safe and comfortable inside.

That could mean machines of a similar size with no cab. It could equally mean a swarm of smaller autonomous machines working on a stand together. The freedom to redesign the machine itself is what I think Generation 3 is really about.

This round takes NFA through Generation 2. Generation 3 looks like a separate, larger build, and I would expect it to come with its own raise when the time comes. Redesigning the machine around the technology is not something NFA can do alone; it has to happen in close partnership with one or more of the big manufacturers. For now the money points at the unglamorous, decisive part: getting a machine to take the right tree, reliably, on every brand of harvester in the field.

Physical AI is long on promise and short on machines that actually do the work. NFA has real machines in real forests, with thousands of hours and paying customers behind them, and the next few years will show whether the promise holds. That is a project worth following.

Axel