April 2025

Dear Investors,

As we enter the second quarter of 2025, the investment landscape remains shaped by uncertainty—from shifting trade policies and elevated tariffs to persistent geopolitical tensions. These factors have raised valid concerns about global supply chains and economic momentum.

Still, we view this period not with alarm, but with resolve. Amid the noise, we see an opportunity to lean into long-term thinking, stay disciplined, and position ourselves for durable growth ahead.

In fact, we believe we are at a historic pivot point in technology – a moment when years of groundwork in automation, artificial intelligence, and electrification are converging to accelerate innovation at an unprecedented pace to create incredible growth opportunities.

In our view, this technological inflection will drive massive value creation in the coming years, far outweighing the transitory macro fears.

Rather than focus on near-term volatility, we dedicate this letter to the deep thematic research—and on-the-ground investigation—that underpins our long-term optimism.

We will explore five powerful trends that we believe are poised to reshape industries and economies:

  1. Tesla’s Full Self-Driving platform and the imminent Cybercab launch: We’re acutely aware that Elon Musk’s behavior has tarnished Tesla’s brand in the eyes of many—and it’s something we’re watching closely. But setting aside the noise, what matters most to us is the company’s core technological trajectory. Tesla’s ambitions in autonomy, particularly its Full Self-Driving platform and the upcoming Cybercab launch, are far more important from an investment perspective. We’re focused on analyzing Tesla’s progress in autonomous driving, its unmatched data advantage, and the immense total addressable market for robotaxis, which we believe could define the company’s future far more than its current public image.

  2. The economic transformation unlocked by autonomy: How self-driving vehicles and autonomy at large could slash transportation costs, create new use cases, and even change the design of cities and consumer behavior.

  3. The rise of humanoid robotics: A deep dive into the companies building general-purpose humanoid robots, with a focus on the accelerating capital flowing into the space, the expanding range of real-world use cases—from manufacturing and logistics to retail and beyond—and the vast, untapped market these technologies are poised to transform.
  1. Warehouse and fulfillment automation – insights from Amazon: What we learned in conversations with former Amazon executives about the current state of warehouse robotics and where it’s heading next.

  2. Autonomous farming and precision agriculture: A field report from our visit to John Deere in Moline, on how autonomous tractors and AI-driven could boost yields, lower costs, and ease labor constraints.

Through these themes, our message is consistent: despite near-term macro fears, the underlying currents of innovation in automation, AI, and energy are stronger than ever.

Ultimately, technology powered productivity gains have driven economic progress for the last 200 years and more. We don’t see that changing. And we continue to believe that we are at the advent of a sea of change through real world AI.

When you look at politically driven or outside shock to markets historically, they occur with regularity. And while they can have an outsized effect in the short-term on any reasonable time frame for long-term investors, they are not what’s material over the long-term. We remain confident, curious, and grounded as we position our portfolio for the potential extraordinary opportunities ahead.

1. Tesla Full Self-Driving and the Dawn of the “Cybercab”

Let’s take a step back.

Investor apprehension often peaks when disruptive visions collide with hard engineering reality – and few visions have been as bold, or as debated, as Tesla’s quest for fully autonomous vehicles. In June however, Tesla’s long-promised autonomy strategy is reaching a critical milestone. Tesla’s team is preparing to launch an autonomous ride-hailing service in Austin, Texas, featuring a dedicated robotaxi vehicle dubbed the “Cybercab.” This is not a distant whitepaper scenario; it’s slated to begin June 2025 as an operational pilot​.

And while this is only the first limited step, for the first time, Tesla vehicles will be rolling through city streets with no human driver in the front seat – a moment that we believe could mark Tesla’s transition from testing self-driving technology to deploying it commercially.

FSD Progress and Tesla’s Data Advantage:

Tesla’s Full Self-Driving (FSD) platform has evolved rapidly over the past year. Internally, Tesla reports that its FSD Beta software (installed in hundreds of thousands of customer-owned cars) is achieving dramatically higher “miles between interventions” – a key metric indicating how far a car can go on autopilot before a human must take over.

While competitors like Waymo have made headlines operating completely driverless cars in limited urban geofences (an important proof of concept), Tesla has been amassing an unparalleled volume of real-world driving data. According to Alphabet, Waymo’s commercial robotaxi fleet drives about 200,000 miles per day autonomously in its service areas, as of February 2025. Tesla’s FSD, by contrast, is already logging roughly 10 million miles per day in autonomous or assisted mode across a far broader range of roads and conditions​.

In other words, even before launching its robotaxi service, Tesla enjoys about a 50x data advantage in real-world driving experience over Waymo​.

This massive scale of driving data – covering myriad edge cases, weather, and road types – acts as a flywheel to improve Tesla’s AI driving models. Every mile provides training feedback.

We continue to believe that data itself is the new gold mine in AI. The rush for novel and useful information can be seen throughout. Encyclopedia Britannica, a company older than the United States has seen a rush towards its data set propelling its business once derided as an antiquity. While this one case — we believe that the more novel and unique (see: autonomous driver edge cases) the more valuable.

Our view is that Tesla is effectively tackling autonomy with the largest real-world driving dataset on Earth, which we believe will only enable more rapid iteration and improvement once true driverless operations begin.

Indeed, Musk has argued that Tesla’s vision-only approach (using cameras without lidar) can work because neural nets trained on billions of miles will eventually handle edge cases as well as or better than dedicated sensors. The upcoming Austin pilot will be the proof point to watch.

It’s worth noting that Tesla’s strategy has shifted subtly but significantly. For years, Musk promised that existing Teslas would receive an over-the-air update and effectively become robotaxis overnight – a “flip the switch” approach. In practice, reaching full Level 4/5 autonomy has been harder than expected, and Tesla is now complementing its software-first strategy with a controlled rollout using a dedicated fleet (much like Waymo and Cruise).

The difficulty of achieving an autonomous, portable, ubiquitous and affordable system has the potential to increase the value of this technology. If it was easy everyone would and could be doing it.

The Cybercab unveiled in late 2024—which we attended—is a purpose-built robo-taxi: a compact two-seater electric vehicle optimized for autonomy and ride-hailing​. It boasts an ultra-streamlined design with only 80 structural parts (versus 200+ in a Model Y), enabling low-cost mass production – reportedly under $30,000 per unit​, according to Tesla in October 2024.

By deploying Cybercabs in one city with geofencing and remote supervision, Tesla can validate full self-driving in a relatively safe setting before scaling up. This is a departure from relying solely on customers to beta-test on public roads; it indicates a maturation in Tesla’s approach to match the caution of competitors. In June, if Cybercabs indeed hit the streets of Austin without drivers, it will be a major milestone – Tesla’s first commercial robotaxi service after years of incremental FSD updates​.

Tesla’s Data and AI Lead:

Why do we believe Tesla is at the cusp of an autonomy breakthrough when many have tried and failed?

The simple answer is data.

This is a classic “tortoise and the hare” story. Waymo sprinted ahead early, but we believe it may be stuck at a local maximum—limited by its reliance on expensive hardware and tightly geofenced maps. Tesla, in contrast, chose a slower but more scalable path. By prioritizing low-cost cameras and mass deployment, it’s building a general solution with far greater long-term potential.

While Waymo uses lidar and HD maps to guide cars in limited areas, Tesla relies on a vision-based system trained by real-world driving data. As of early 2025, more than 4 million Teslas with Autopilot/FSD hardware are on the road, constantly collecting footage and telemetry. No other company—even all AV firms combined—comes close to that scale. As one former Lyft autonomy exec told us in a recent In Practise interview:

“Data is crucial, especially now when we look at the big models, even in autonomous driving. Consider the foundational models from OpenAI, Google, and others. They have utilized almost all available data from the Internet or that they can purchase. However, they still struggle with tasks involving street scenes.” Source: This interview is featured on the In Practise Platform, a primary research service .(Source: https://inpractise.com/articles/tesla-vs-waymo-autonomous-driving-and-long-tail-data).

All In Practise interviews are for informational purposes only and should not be relied upon as a basis for investment decisions. In Practise is an independent publisher and all opinions expressed by guests are solely their own opinions and do not reflect the opinion of In Practise or Nightview.

In other words, Tesla’s huge fleet effectively crowdsources the solution to self-driving by learning from millions of human decisions and mistakes on the road. This data advantage does not guarantee success (it must still be translated into robust driving policy by AI), but it creates a high barrier to entry. If Tesla’s FSD AI can reach human-level reliability, we believe it can immediately be deployed to a fleet orders of magnitude larger than any competitor’s​.

That scaling potential is one reason we remain bullish on Tesla’s approach.

Current Progress – Are we there yet?

While in the immediate term we  approach some of Tesla’s autonomy claims with measured optimism. FSD (Supervised) v13.2.8 for Hardware 4 (HW4) has steadily improved. And we believe the rate of improvement validates the company’s approach.

Users report the latest builds handle complex urban maneuvers much more confidently than a year ago – unprotected left turns, roundabouts, and even some corner-case interactions like debris on the road.

While Musk’s timelines are famously optimistic, by Q1 2025 Tesla indeed felt confident enough to schedule the unsupervised Cybercab launch. Notably, Tesla is adding a layer of remote human oversight for safety – they have hired teleoperators who can monitor the robotaxis and intervene via remote control if a vehicle gets confused. (Source.)

This “unsupervised but not unmonitored” approach is prudent, in our view. It acknowledges that true general AI-level driving may still falter occasionally, but those edge cases can be managed with a human backstop.

In essence, Tesla is doing what Cruise and Waymo do (remote fleet supervision), but with a far larger pool of situations to learn from. If the Austin pilot goes well – i.e., minimal disengagements and no serious incidents – it could unlock a path to scale this service to other cities and eventually eliminate the remote monitor step.

Total Addressable Market (TAM) for Autonomous Ride-Hailing:

The economic opportunity behind Tesla’s robotaxi push is enormous. By turning cars into software-driven revenue generators, Tesla aims to transform the business model of transportation. Instead of selling a car once, a robotaxi can potentially earn tens of thousands of dollars per year by providing transportation as a service.

Importantly, the total addressable market in terms of consumer spend on global transportation is even larger – our calculations suggest trillions of dollars of economic redistribution could be credited to a shift towards autonomy—which we wrote about here in Q3 2024.

This is why Tesla – and we as investors – see autonomy as a multi-decade value driver. What could this market be worth to Tesla specifically? In a recent earnings call, Elon Musk has boldly predicted that Tesla’s robotaxi network, combined with its humanoid robots (more on Optimus later), could make Tesla the most valuable company in the world. In a recent earnings call.

And while our own estimates are more conservative, but still staggering: even a few percent share of global miles shifting to Tesla robotaxis in the late 2020s could yield billions in incremental operating profit.

Cybercab Launch – What to Watch:

In June, Tesla’s Austin pilot will begin with a limited service area and a fleet of company-owned vehicles (initially, possibly modified Model Ys and early Cybercab prototypes). As the service scales commercially, we expect the long-run cost curve to trend downward as utilization increases and human drivers (and their labor cost) are removed.

Tesla has a huge margin opportunity here: thanks to EVs’ lower operating costs (about one-third the cost of gasoline vehicles) and expected high utilization, Tesla could undercut competitors while still earning a healthy “take rate” on each trip​.

The Cybercab rollout will also test Tesla’s safety and public trust. The company claims that even in beta form, FSD-equipped cars have accident rates far lower than average.

In summary, Tesla’s FSD platform is entering a prove-it period.

The groundwork – in data, AI, and vehicle design – has been laid to enable a true driverless service. Investors often ask us, “When will autonomy really happen at scale?” Our answer is that the pivot is happening now.

Like the rollout of the internet or smartphones, adoption may seem gradual and then sudden. We expect Tesla to start small in Austin, gather feedback, then rapidly iterate its software. If all goes well, by 2026 Tesla could be operating autonomous fleets in multiple U.S. cities.

We believe the Total Addressable Market (TAM)  is in the trillions, the competitive moat is significant (data, vertical integration, unit economics), and the consumer value proposition (cheaper, safer transport) is extremely compelling. This is why we remain long-term believers in Tesla and view 2025 as potentially the year Tesla “cracks the code” on commercial self-driving.

In our view, it’s the dawn of the Cybercab era – and potentially the start of a transformation in how humans move around.

2. Autonomy’s Economic Transformation: Low-Cost Transport and New Use Cases

Stepping back from Tesla specifically, let’s consider the broader implications of ubiquitous autonomy.

We often talk internally about the deflationary effect of technology – how advances drive down costs and unlock new demand. Autonomous vehicles (AVs) are a prime example. By automating the driver, we are poised to radically reduce the cost per mile of transportation, perhaps to levels never seen in modern times. This is not just about Uber fares getting cheaper; it’s about fundamentally changing what is possible in logistics, urban planning, and consumer behavior.

Cost-Per-Mile Revolution: Today, owning and operating a personal car costs around $0.80 per mile in the U.S. (when factoring fuel, insurance, depreciation, etc.), and hiring a ride (taxi or ride-share) often costs $1-$3 per mile, depending on the geography.​

These costs have kept certain use cases in check – for instance, few people would take a 50-mile daily commute by taxi, or have food delivered from a restaurant 30 miles away, because the economics don’t make sense.

Autonomy flips this equation on its head. Without a human driver to pay and with optimal (e.g. cars being in use potentially 24 hours a day as opposed to a fraction of that) utilization, the cost of an electric robo-taxi could drop to mere pennies per mile above electricity and maintenance expenses. Tesla has floated numbers around $0.30-$0.40 per mile to the end user for a robotaxi ride​, for instance.

When transportation becomes that inexpensive, we use more of it. Lower price points stimulate demand – a well-known economic effect. For instance, if a trip across town costs $3 instead of $15, people will take more trips, and they’ll be willing to travel farther for the same activities.

What might this look like in daily life?

We could see new use cases bloom: families might opt to not own a second car, instead relying on robo-taxis for occasional needs (thus expanding the ride-hail market). Individuals might live further from city centers if commuting in an autonomous pod lets them work or relax during the ride – turning dreaded traffic time into productive or leisure time.

On-demand mobility could reach communities and demographics currently underserved by transit. Even something as simple as running errands could change: your autonomous car could drop you at the store entrance and go park itself or go pick up another family member, then return when summoned – saving time and parking hassle.

Changes in Logistics and Delivery:

Autonomy isn’t just about moving people. Moving goods autonomously is equally transformative. We are starting to see pilot programs for autonomous delivery vans, sidewalk robots, and even drones for last-mile delivery.

The economics of e-commerce could shift further in favor of delivery over in-person shopping when human labor is less of a limiting factor. For instance, today a significant cost of grocery delivery is the driver’s time; autonomous delivery bots or vehicles could make it feasible to get a pint of ice cream delivered at midnight for a trivial fee.

The trucking industry, faced with chronic driver shortages, is also eager for solutions.

Autonomous trucks could operate almost nonstop (only pausing for fuel/charge and maintenance), drastically shortening delivery times cross-country. A truck that can drive 20+ hours a day could nearly double the output of today’s truck that is limited by human hours-of-service rules. This means faster shipping and potentially lower freight costs, which in turn lowers costs of goods for consumers. And ultimately cheaper goods and services.

Rethinking Consumer Behavior:

As autonomy reduces the “cost” of distance – both in money and in mental effort – consumers may change behaviors in subtle ways. We might see more cross-town dining and entertainment if getting there is cheap and easy. Package returns or store pickups (think curbside) become more convenient when your car can run errands without you.

Families could send an autonomous vehicle to shuttle kids to activities safely, without a parent having to leave work (a concept many parents would welcome!). Car interiors might evolve to be more like offices or lounges since occupants are not driving; this could spawn an ecosystem of in-car services and entertainment.

In sum, mobility on demand may start to resemble an app utility, blending into our lives in ways we don’t even fully anticipate yet – much like how the smartphone’s always-on connectivity led to new habits (constant communication, on-demand everything).

Steve Jobs famously introduced the iPhone as a “an iPod, a phone, and an internet communicator” but even with his ambitious vision he fell short of the impact. It created entirely new industries, ride hailing among them, we see a similar potential here.

The common thread: when the price drops and convenience rises, usage expands.

Economic Productivity and Inclusion:

On a macro scale, widespread autonomy could boost productivity.

Consider commuting: millions of hours are spent driving, which is essentially dead time in terms of work output (and often not very relaxing either). If those hours are freed up – people can work, read, or rest during commutes – that’s a gain for productivity or well-being. Moreover, mobility-as-a-service could serve populations unable to drive (elderly, disabled) far better than current options, increasing their access to commerce and social life.

We could see higher labor force participation in some segments if commuting or transportation is no longer a barrier (imagine someone visually impaired gaining independent mobility via AV services).

From Autonomous Cars to Autonomous Everything:

The lessons and technologies of self-driving cars (sensors, AI decision-making, robotics) are spilling over into other domains.

We’re entering an age where autonomy will not be limited to cars. Drones are achieving autonomy in the skies (for inspection, delivery, agriculture monitoring). Warehouse robots (more on that in the next section) autonomously ferry goods within fulfillment centers. Factories are increasingly using autonomous mobile robots to transport parts. Even in software, “autonomy” in the form of AI agents could handle tasks for us without constant direction.

The common thread is AI taking over repetitive or dangerous tasks, and working tirelessly. This has broad economic implications: it could lead to higher output with less input, which is the definition of productivity growth. Some worry about job displacement, and indeed certain job categories (e.g. taxi and truck drivers) will [MC1] eventually decline.

While there is a chance this time is different, always dangerous words, the Luddites have been consistently wrong. History shows technology creates new jobs even as it displaces old ones – often jobs that are higher skilled or entirely new industries. We believe the net effect of autonomy will be to boost economic growth.

3. Rise of the Humanoid Robots

If autonomous cars represent AI seeping into how we move around, humanoid robots represent AI coming to life in a physical form to help us with work.

This is no longer science fiction.

Over the last 18 months, we’ve seen a proliferation of prototypes and startups focused on general-purpose, human-like robots. Tesla’s bipedal robot Optimus (also known as the Tesla Bot) grabbed headlines when it was first revealed in 2022, but it’s far from alone now.

Companies like Figure AI, Agility Robotics, Apptronik, Sanctuary AI, and others have all made significant strides with humanoid designs. Many are chasing the ultimate holy grail of products a workable, intelligent, and economic labor supply. We are witnessing the dawn of a new industry – one that aims to build robots that can work in our factories, warehouses, and even offices and homes someday.

In Q1 2025, Nightview’s research team did a deep dive on this emerging field, attended demos and spoke with former executives at some of these companies. Our goal here is to understand who’s leading, what use cases are emerging first, and how big the opportunity might be.

First, let’s clarify why so many are pursuing a humanoid form factor.

The human environment (buildings, tools, infrastructure) is designed around the human shape and abilities. A robot that can eventually navigate our world with similar agility and use the same tools as we do (handle a drill, carry a box, climb stairs, etc.) has a near-infinite range of potential tasks.

Rather than designing a special robot for every job (one for welding, another for dishwashing, another for elderly care), a general-purpose humanoid could, in theory, be trained or instructed to do many different tasks, much like a person can learn new skills.

This is the long-term vision: “robot workers” that can take over mundane or hazardous jobs and augment human labor in countless ways. It sounds fantastical, but so did self-driving cars at one point – and now we see those on the cusp of reality. The pace of progress in robotics (thanks to advances in AI, sensors, and actuators) has accelerated, and importantly the cost to build these robots is coming down dramatically (as noted, cheaper components and better designs).

Even under more conservative assumptions, the TAM is huge; for example, even a 10% penetration of manufacturing tasks by humanoids would imply hundreds of billions in robot sales. And that’s just on a replacement level of current labor we ultimately see this as *increasing* the overall workforce dramatically.

That’s the opportunity.

4. Automating the Warehouse: Insights from Amazon’s Robotics Efforts

While autonomous cars and humanoid robots capture imaginations, a more quiet revolution has been underway in warehouses and fulfillment centers for years. Nowhere is this more evident than at Amazon, a company that has become as much a robotics and logistics firm as it is an e-commerce giant.

In Q1, we had the opportunity to speak with former Amazon executives and engineers who were instrumental in building Amazon’s warehouse automation programs. Their firsthand insights provided a vivid picture of how far automation has come inside Amazon’s operations – and how far it will go in the near future.

Here, we share some key takeaways, along with data on Amazon’s current robotics fleet and what’s next.

From Kiva to Proteus – The Robotics Fleet Today:

Amazon’s journey in warehouse automation kicked into high gear in 2012 when it acquired Kiva Systems. Kiva’s orange robot units famously scurry under shelves and move them around to bring items to human pickers. Fast forward to 2025: Amazon now has an army of over 750,000 robots working across its fulfillment and sortation centers.

This is a staggering number – more than double what they had just a few years ago​ – and makes Amazon the world’s largest commercial user of mobile robots. According to Amazon’s own disclosures, these robots come in at least 8 different types​:

  • Automated guided vehicles (AGVs) like the original Hercules and Pegasus drive units that ferry shelves (pods) to packing stations​. These handle the bulk of item movement in older Amazon fulfillment centers.

  • Robotic arms: e.g., Robin and Cardinal, which can pick up and sort packages and envelopes at high speed. Robin grabs packages off conveyors and places them on carts or other conveyors​, while Cardinal can pluck a specific parcel from a pile, read its label, and reroute it – handling packages up to 50 lbs​. These reduce manual sorting work.

  • Sparrow, a newer robotic arm unveiled in late 2022, is perhaps the most advanced – it uses AI and computer vision to identify individual items in bins and pick them up one by one​. This tackles the holy grail of “each-picking,” the task of picking out single products for orders. Initially Sparrow could handle something like 65% of Amazon’s product assortment (shapes it can pick); Amazon is working to expand that capability. By moving items into totes, Sparrow preps them for packing with far less human labor.

  • Proteus, launched in 2022, is Amazon’s first fully autonomous mobile robot that can navigate unrestricted (not confined under shelves)​. Proteus can roam the warehouse, using sensors to avoid people and obstacles. Its job currently is to carry carts of packages (like “GoCarts”) from sorting areas to loading docks, working in tandem with the Cardinal arm that fills the carts​. Proteus represents a shift: robots that can work in the same space as humans, rather than in caged-off robot-only zones.

  • Other systems like Sequoia (an inventory storage system using mobile robots to stow goods more densely)​, and automated packing machines that size and bag orders efficiently​, round out the robotic fleet.

One Former Head of Systems and Products at Amazon Robotics we spoke with highlighted how these systems interlink:

When you think about labor in a warehouse, the task that takes the most time is walking. Amazon has largely addressed this with Kiva and Proteus. The next biggest cost for humans is physically manipulating items. Unlike a cuboidal cardboard box, items like an iPhone or a water bottle have very different form factors, making them challenging to pick up. The orientation of these items is always unpredictable. Part of the brilliance of the Amazon inventory system is random storage, which enables efficiencies but introduces complexities for automation. A tote might arrive with 15 different items, and you never know which 15. They could be 15 different items, or it could be four of one, five of another, and three of something else. (Source: This interview is featured on the In Practise Platform, a primary research service https://inpractise.com/articles/amazon-robotics-sparrow-sequoia-and-automation-economics)

All in Practise interviews are for informational purposes only and should not be relied upon as a basis for investment decisions. In Practise is an independent publisher and all opinions expressed by guests are solely their own opinions and do not reflect the opinion of In Practise or Nightview.

In a modern Amazon fulfillment center, a customer order might be processed with minimal human touches: robots bring the shelf of products, a human (or Sparrow in increasing cases) picks the right item, it’s conveyed to packing where a machine or person packs it, then Robin or Cardinal sorts the package, Proteus takes the container of sorted packages to the dock, and humans do the final truck loading (for now).

Each step that transitions to automation increases efficiency. Amazon reported that its newest generation facilities, like the one in Shreveport, Louisiana, which is loaded with all these latest robots, is seeing about 25% improvement in productivity versus older centers​.

That is a huge gain in an industry as optimized as warehousing. A 25% boost means faster delivery, lower cost per unit, or the ability to handle more volume with the same staffing.

Worker Augmentation, Not Replacement (Yet):

A point emphasized by the Amazon veterans is that the goal has been augmenting human workers, not eliminating them. Amazon’s internal mantra has been to make “employees’ workdays safer and more productive” through automation​.

For example, robots fetching items means workers walk less – a single picker might have walked 10+ miles a day pre-Kiva, but now can stand at a station and have items brought to them.

Likewise, lifting heavy packages is increasingly done by robotic arms or lift assists, saving human backs. Amazon publishes safety data indicating significantly fewer injuries in sites with more robotics, and an MIT study noted in Amazon’s news release found 60% of employees felt positive about the impact of robotics on their jobs​.

Of course, Amazon’s workforce has continued to grow even as robots deployed (hundreds of thousands of employees have been added since 2012 in operations). The company is handling far more package volume than it could without automation. The former execs predict that in the foreseeable future, humans and robots will work side by side, with robots handling the grunt work – fetching, carrying, sorting – and humans handling exceptions, delicate tasks, and final touches.

However, they acknowledge the balance is shifting. As technology matures, robots will take on more. Ten years ago, robots in warehouses only moved shelves. Now they sort outbound packages and even pick items. The next ten years could bring near-fully autonomous warehouses, albeit with human supervisors.

The role of AI and simulation:

Amazon’s robotics team leverages a ton of AI in simulation to test robots.

Before Proteus ever roamed a live warehouse, it likely ran millions of virtual scenarios to ensure its navigation software was solid. Similarly, the vision systems for Sparrow were trained on models of Amazon’s products.

One tidbit we learned: Amazon’s vast product catalog actually helped them here – they have images of most items sold, which can train an AI to recognize and plan grasps for each item. It’s like teaching Sparrow the shape and weight of an item before it ever “sees” it in a bin. This reduces errors. In essence, Amazon can use its data at scale (catalog data, order patterns, etc.) to inform the robots’ brains.

Logistics Chain Automation:

It’s not just the fulfillment center.

Amazon is automating across inbound logistics (unloading trucks with robotic arms and conveyors), middle-mile (they have sortation centers with similar tech to group packages by destination), and last-mile (delivery).

On delivery, while they experimented with sidewalk robots (“Scout”) and have patents for delivery drones, the biggest impact has come from routing software and an app that effectively automate a driver’s workflow (telling them which package to deliver next, optimizing routes in real-time).

Fully driverless delivery vehicles are still being tested (Amazon is an investor in Rivian, and one could envision a future where some Rivian electric delivery vans get autonomous capabilities in limited areas).

For now, Amazon’s focus is on squeezing time and cost out of the warehouse to delivery pipeline. Amazon’s obsession with speed (Prime same-day delivery expanding) is directly enabled by these robotics innovations and machine-learning capabilities.

As one Former Robotics Director at Amazon told us:

“One of the significant productivity advancements during my time there was the use of machine learning to reduce the number of items requiring barcoding. Although the job remains the same, machine learning was employed to ensure, with high confidence, that the right item is placed in the correct location. This is the stowing part, which is rarely shown online due to its messy nature, involving packaging and dunnage. (Source: This interview is featured on the In Practise Platform, a primary research service https://inpractise.com/articles/the-rise-of-warehouse-automation-and-amazon-robotics-journey)

All in Practise interviews are for informational purposes only and should not be relied upon as a basis for investment decisions. In Practise is an independent publisher and all opinions expressed by guests are solely their own opinions and do not reflect the opinion of In Practise or Nightview.

Scale and Metrics:

The scale at which Amazon operates makes even incremental improvements massive. With 750k+ robots and perhaps 1+ million human employees globally, Amazon fulfillment ships ~10 billion items a year, as of last year.  

So a 1% efficiency gain can mean 100 million more items shipped for the same cost.

Here are a few concrete data points:

  • Amazon’s robotic drive units (like Hercules) can carry shelving pods that weigh up to ~3,000 lbs. They navigate by reading unique QR codes on the floor grid for positioning​. Over the years they’ve improved speed and reliability, reducing downtime.

  • Robin arms can handle upwards of 300+ packages per hour each, and Cardinal arms can process roughly 1,000 packages per hour in sortation, far faster than a human could sort by hand. Cardinal also reduces injuries from lifting heavy boxes by automating that​.

  • The new Sequoia system Amazon launched can stow inventory 75% faster than previous methods​. It uses mobile robots to dynamically re-slot inventory, effectively compressing space and bringing items to workers more efficiently.

  • Overall, Amazon stated the Shreveport facility uses “8 different robotics systems working in harmony” and is a model for the future​.

Work left to be done:

Our conversations revealed there is still more automation left in the pipeline that we believe will make fulfillment more efficient—and ultimately drive incremental operating margins.

A few key areas:

  • Automating inbound receiving – unloading trailers is still manual; Amazon has been testing robotic arms that can pick up random items from floor-loaded trailers (a very challenging task due to variety of shapes). This could significantly speed up how inventory gets into the system.

  • More autonomy in last-mile – while human drivers will be around for a while, expect more driver-assist tech in Amazon’s fleet, maybe even some autonomous delivery pilot in a sunny suburb (perhaps in partnership with an AV company).

  • Data, data, data – Amazon will leverage AI on all the data these robots generate. For instance, analyzing how frequently items are picked together to store them optimally (that’s already done) or predictive maintenance on robots to avoid downtime.

  • Scaling up human-robot training – As more robots join the workforce, Amazon spends a lot on training employees to work with them. They have “Robotics Safety and Training” programs. In the future, they might even have employees remotely monitoring multiple sites – akin to air traffic control for robots – which ties back to our point about new types of jobs created.

From an investment perspective, we believe Amazon’s lead in warehouse automation solidifies its competitive moat in e-commerce. It’s hard for smaller competitors to match the efficiency (Amazon’s fulfillment cost per unit keeps coming down, which can translate to better prices or faster delivery).

That said, the technologies Amazon develops often spread industry-wide eventually (e.g., Kiva robots are now used by others as similar systems are sold by various vendors).

We see opportunities in companies supplying advanced sensors, AI software, or specialized machinery to the likes of Amazon, Walmart, and other retailers automating their logistics. Moreover, Amazon’s success creates a blueprint for others – for example, we observe companies like Walmart and Target also heavily investing in warehouse automation (Walmart has added robotics to many of its regional distribution centers, and Target purchased sortation robots).

The entire retail ecosystem is moving this way to stay competitive.

In conclusion, Amazon’s fulfillment operations give us a window into the future of work.

The warehouse of the future is arriving, one robot at a time – and Amazon’s experience gives us confidence that efficiency gains from automation are real and material. Just as importantly, it shows that robots and humans can co-exist productively, each doing what they’re best at. In an economy grappling with labor shortages and inflation, that’s a very encouraging story.

5. Field Report: John Deere – Autonomous Tractors and Precision Agriculture in Action

In February, our team took a trip—arguably ill-timed when it was -10 degrees)—to Moline, Illinois – home of John Deere – to get a close look at how autonomy and AI are revolutionizing agriculture.

John Deere has been aggressively integrating advanced technology into farm equipment, effectively turning tractors and combines into high-tech robots of the field.

We spent time with Deere’s CTO and CFO, saw demos of autonomous tractors, and spoke to former execs and customers using Deere’s precision ag tools.

The result was a deeper appreciation for how automation is boosting yield, cutting costs, and addressing the acute labor challenges in farming.

Here are the key insights and data points from that field visit.

Autonomous Tractors: From Concept to Reality –

A highlight of our visit was learning about Deere’s latest autonomous tractor, a retrofitted 8R series tractor equipped with a suite of cameras and sensors.

This is the same system Deere first unveiled at CES 2022. It uses six pairs of stereo cameras for 360-degree vision and advanced AI models to navigate fields without a driver. The machine is hooked up to a chisel plow, performing tillage on a test field – completely on its own.

It carefully stayed in the bounds of the field, detected and stopped for obstructions (we saw it identify a stray log lying in the path), and sent real-time video and alerts to the farmer’s phone. We were struck by how “normal” the tractor looks – aside from the camera pods, there’s no flashy antenna or anything; the brain is largely invisible.

Deere is not stopping at the 8R row-crop tractor. At CES 2025 (this January), the company announced a second-generation autonomy platform and plans to extend it to more types of equipment: smaller tractors for orchards and vineyards, and even construction machines like articulated dump trucks for quarries. The CTO, Jahmy Hindman, whom we met briefly, emphasized that we are in early innings for development.

Addressing Labor Shortage & Efficiency:

Agriculture is facing a well-documented labor crunch. The average U.S. farmer is nearly 60 years old, and younger people are less inclined to do strenuous farm work​. By one estimate, in 2024 the U.S. AG industry had 2.4 million fewer workers than needed​, and this gap is only growing.

We heard anecdotes from farmers about advertising tractor operator jobs with little uptake. Autonomous tractors are arriving just in time to fill this gap. This highlights a key advantage: 24/7 operation. An autonomous tractor doesn’t need to rest. It can prepare fields overnight while the farmer sleeps.

Precision Agriculture: Data-Driven Yield Improvements:

Autonomy in driving is one aspect; precision agriculture (using data and AI to optimize farming) is another area where Deere is leading, and the two go hand in hand. We delved into Deere’s See & Spray™ technology, which is an AI-driven system that identifies weeds in the field and targets them with herbicide, instead of blanket spraying the whole field.

The results are astounding: in the 2024 growing season, farmers using See & Spray reported on average 59% less herbicide use on their corn, soybean, and cotton fields​. In total, Deere calculated about 8 million gallons of herbicide mix saved across 1+ million acres using this tech in 2024​.

Conclusion: Our Core Belief About the Future

Across all these themes – from robo-taxis in city streets, to humanoid robots in factories, to autonomous machines on farms – we believe one narrative is clear: technology is compounding to drive an era of extraordinary innovation, value creation and economic growth.

Yes, the world is grappling with short-term challenges. Tariffs, geopolitical rifts, inflationary pressures – these can disrupt markets in the near term.

But as long-term investors, we take solace (and build conviction) in the fundamental trajectory of progress. The underlying trends of automation, artificial intelligence, and the energy transition (electrification and sustainable energy underpin much of these advances) are not only intact, they are accelerating.

We are truly at a historic pivot point.

In our opinion, the 2020s are shaping up to be akin to the dawn of the internet in the 1990s or the mobile revolution in the 2000s – a time when bold bets on disruptive technology can pay off massively.

In this letter, we’ve highlighted how Tesla is on the verge of potentially launching a transformative autonomous transport network, how autonomy could reshape economies, how a new industry of humanoid robotics is being born, how Amazon’s automated warehouses are redefining retail logistics, and how John Deere’s smart machines are boosting agricultural productivity.

In each case, these are examples of long-term thinkers executing through short-term noise. Tesla didn’t abandon FSD development even when critics howled – now they might be first to deploy at scale.

Amazon invested in robots for a decade – now it has a moat competitors struggle to match. John Deere methodically built precision tech for 20 years – now they’re solving farming problems others can’t.

Our core belief is that the exponential advancement of technology will continue to unlock outsized growth for those who position themselves ahead of the curve.

We remain confident that companies at the forefront of automation and AI will create enormous shareholder value. There may be volatility along the way – breakthroughs are not linear and market sentiment can swing – but the end-state we envision is one of significant wealth creation and positive impact.

We see a future where transportation is safer and far cheaper (benefiting society and yielding profits for platform operators), where robots elevate productivity and take on drudgery (improving margins for businesses and quality of life for people), and where clean energy and intelligent systems make industries from manufacturing to agriculture more sustainable and efficient.

Investing in this future requires patience and imagination. There will be skeptics at every turn, and not every bet will pan out.

But as we’ve seen even in just this past quarter’s developments, the momentum is undeniable. Our job is to be judicious – to separate hype from viable progress – and to allocate capital to the most promising opportunities with a long-term horizon.

We want to thank you, our LPs and shareholders, for sharing this vision and for your continued trust. It’s your capital that allows us to support and partake in these transformative ventures.

Despite the choppiness of the current macro environment, we encourage you to focus on the bigger picture outlined above. The companies delivering real innovations are, in many cases, improving their competitive position even as the market worries about the latest headline.

Eventually, true value shines through.

In closing, we reiterate our conviction: we are on the cusp of an automation and AI-driven renaissance.

The coming years, from our perspective, will likely surprise to the upside in terms of technological achievements and their economic ramifications. We remain confident, curious, and grounded – confident in the opportunities, curious to discover the next great ideas, and grounded in rigorous research (as we’ve shared here) to guide our decisions.

Thank you for your partnership and support.

We look forward to navigating this exciting future together.

Disclosures

The opinions expressed herein are those of Nightview Capital and are subject to change without notice. The opinions referenced are as of the date of publication, may be modified due to changes in the market or economic conditions, and may not necessarily come to pass. Forward-looking statements cannot be guaranteed.

This is not a recommendation to buy, sell, or hold any particular security. There is no assurance that any securities discussed herein will remain in an account’s portfolio at the time you receive this report or that securities sold have not been repurchased. It should not be assumed that any of the securities transactions, holdings or sectors discussed were or will be profitable, or that the investment recommendations or decisions Nightview Capital makes in the future will be profitable or equal the performance of the securities discussed herein.

Nightview Capital reserves the right to modify its current investment strategies and techniques based on changing market dynamics or client needs. Recommendations made in the last 12 months are available upon request.

This article contains links to 3rd party websites and is used for informational purposes only. This does not constitute as an endorsement of any kind. While Nightview uses sources it considers to be reliable, no guarantee is made regarding the accuracy of information or data provided by third-party sources.