In 2012, Amazon founder Jeff Bezos was asked by TV host Charlie Rose whether his e-commerce company would ever venture into brick-and-mortar stores. Bezos said shoppers were well-served by existing physical retailers and that Amazon wasn’t interested in launching a “me-too” product.
“We want to do something that’s uniquely Amazon,” Bezos said. “If we can find that idea, and we haven’t found it yet, but if we can find that idea, we would love to open physical stores.”
Six years later, Amazon landed on a revolutionary retail concept that it hoped would transform how people shop in brick-and-mortar stores. The company launched its first Amazon Go convenience store featuring a new kind of technology, called “Just Walk Out.”
In practice, customers would be able to load up their cart and exit the store without standing in a checkout line. Amazon soon brought cashierless checkout to its Fresh supermarkets and two Whole Foods locations. In 2020, the company began licensing Just Walk Out technology to third parties, signing on retailers in stadiums, airports and hospitals.
But the company has since taken a sideways turn.
In April, Amazon announced it was removing cashierless checkout from its U.S. Fresh stores and Whole Foods locations, a move that coincided with CEO Andy Jassy’s efforts to rein in costs to meet rapidly changing macro conditions.
In place of Just Walk Out, which typically requires ceiling-mounted cameras, shelf sensors and gated entry points, Amazon Fresh stores and Whole Foods supermarkets will feature Dash Carts. The carts track and tally up items as shoppers place them in bags, enabling people to skip the checkout line. Amazon continues to use Just Walk Out in its grab-and-go marts and UK Fresh stores.
A woman uses a dash cart during her grocery-shopping at a Whole Foods store as Amazon launches smart shopping carts at Whole Foods stores in San Mateo, California, United States on February 25, 2024. The smart shopping cart makes grocery shopping quicker by allowing customers to scan products right into their cart as they shop and then skip the checkout line.
Tayfun Coskun | Anadolu | Getty Images
The main challenge for Amazon and other startups working on autonomous checkout is the need to scale it to enough locations and retail categories that it becomes a natural part of in-store shopping, said Jordan Berke, founder and CEO of retail consulting firm Tomorrow.
“Until that’s the case, it’s an uphill battle,” Berke said. “These technology providers, Amazon included, are going to have to subsidize and continue to invest to train the retailer, train the consumer, train the market, that this is a mainstream experience that we can all trust and not need to think about as we walk in and out of a store.”
‘The hardest problem to solve’
At one point Amazon saw Just Walk Out becoming a core part of the experience of shopping in its physical stores. The company in 2018 planned to open as many as 3,000 Amazon Go stores within a few years, Bloomberg reported at the time, citing people familiar with the plans.
Bezos had assigned top talent from across the company, including a longtime Amazon executive who built the original Kindle e-reader, to work on cashierless checkout. The technology was considered a key ingredient in Amazon’s long-running pursuit to become a giant in the $1.6 trillion U.S. grocery market.
When Amazon debuted Just Walk Out in January 2018, it was a “quake moment” for the industry, causing Walmart and “almost every other retailer” to leap into action and consider developing their own vision-based checkout systems, said Berke, who previously led Walmart’s e-commerce business in China.
Amazon and other retailers soon learned that automating the checkout process is “the hardest problem to solve,” Berke said. Cashierless checkout systems require a hefty upfront investment to blanket a store with overhead cameras and hire staff to label and review shopping data.
“It meant a store had to dramatically increase its sales in order to pay off that investment,” Berke said.
Walmart teams found as part of a cost analysis in early 2019 that it would run a retailer between $10 million and $15 million to create a similar computer vision-based checkout system for a 40,000 square foot supermarket, Berke said.
Just Walk Out became an expensive project for Amazon, too. In 2019 and 2020, the company shelled out roughly $1 billion per year, including research and development costs and capital expenditures, to “learn and scale” the technology, Berke said. He said those figures are based on discussions with a former Just Walk Out executive who left Amazon to join Walmart.Amazon didn’t provide a comment on the figures.
Many retailers have since moved on from computer vision in favor of simpler methods like mobile checkout through an app, Berke said.
While it’s no longer featuring Just Walk Out as prominently in its own stores, Amazon says it has inked deals with a growing list of customers. More than 200 third-party stores have paid Amazon to install the cashierless system. The company expects to double the number of third-party Just Walk Out stores this year, Jon Jenkins, who previously served as vice president of Amazon’s Just Walk Out technology, said in a recent interview. Jenkins departed Amazon in late September to become technology chief of electric bike and scooter startup Lime, according to his LinkedIn page.
Jon Jenkins, Amazon’s former vice president of Just Walk Out technology, gives a tour of the mock convenience store where the company tests its cashierless checkout system in Seattle, Washington, on August 22, 2024.
CNBC
Jenkins disputed characterizations that Amazon’s phasing out of Just Walk Out from its own supermarkets represents a setback or a sign of the technology’s demise. He said Amazon proved through tests in its own grocery stores that the technology is “incredibly capable,” noting it deployed the system in large supermarkets with “600 people in the store at the same time.”
Other startups such as AiFi and Grabango have developed autonomous systems for supermarkets, convenience stores and other retailers, but widespread adoption has been slow, as the technology remains costly and challenging to operate in large store formats.
Inside the lab
Amazon is still fine-tuning its Just Walk Out technology.
In August, CNBC got the first on-camera look at a mock convenience store where Amazon tests the system before deploying it in third party retailers and its own stores.
The testing lab, which it calls “beverage base camp,” is located in Amazon’s Seattle headquarters. It has faux gates that mimic the experience of scanning your smartphone or credit card to enter a Just Walk Out store. The walls are lined with shelves of typical grab-and-go products like Milky Way bars, pita chips and gum, and there are coolers stocked with Coke cans and other beverages.
Amazon sets up Just Walk Out stores by first creating a 3D scan using LiDAR machines or iPads that help it determine where to place cameras so they have the clearest view.
“The goal is to have the fewest number of cameras possible, so we optimize the camera placement so that we can get enough coverage on each fixture to see what is happening in the store,” Jenkins said.
The system determines what shoppers purchased using several inputs, including the 3D scans, a catalog of product images, the video footage, and weight sensors on the shelves. Amazon in July updated the AI system behind its Just Walk Out technology to handle all the inputs in a store simultaneously.
The new “multi-modal” system can generate receipts faster by more accurately predicting which items shoppers have picked up and put back on shelves. The company said these changes should make it “faster, easier to deploy and more efficient” for retailers who install the system in their stores.
Amazon’s “primary focus” is selling the technology to third-party businesses and deploying it in small to medium-sized store formats, where the system “tends to generate a little better [return on investment],” Jenkins said. Earlier this year, Amazon also began selling its connected grocery carts to third parties.
Amazon in September announced several new third-party Just Walk Out stores at universities and sports stadiums.
CNBC
At one Just Walk Out store, inside Seattle’s Lumen Field, home to the NFL’s Seahawks, the company said it boosted sales by 112% last season, with 85% more transactions during the course of a game.
“It was awesome that we had our own stores as the laboratory to sort of build and launch this,” Jenkins said. “But over time, like many things at Amazon, the success of this project and the product will depend on third parties adopting the technology. There will always be more third-party stores in the world than there will be first-party stores.”
Amazon has used a similar playbook in in the past. Amazon Web Services, the company’s wildly successful cloud-computing unit, originated from the company’s need for IT infrastructure to support its fast-growing online retail business. And in recent years, Amazon has leveraged its logistics and fulfillment network to provide services for third parties.
With Just Walk Out, Amazon faces the challenge of convincing retailers that they can trust one of their biggest competitors with handling valuable shopper data.
In 2022, Amazon moved the team behind Just Walk Out from its retail organization to AWS. It marked one of the clearest signals yet that Amazon is serious about selling the technology to other retailers, and could help ease some fears among rivals.
“They’re clearly in sales mode,” said Sucharita Kodali, retail analyst at Forrester Research, in an interview.
Kodali said Amazon still has a “long way to go” before the technology is ubiquitous. Getting there will require patience from Amazon investors and data that shows both retailers and shoppers are embracing the technology.
“There’s almost a viral effect that will occur over time,” she said. “It’s just going to take a long time because you’ve got to cycle through everybody in America having this experience, and for the most part, it’s just Amazon fighting this fight right now.”
Neurodiverse professionals may see unique benefits from artificial intelligence tools and agents, research suggests. With AI agent creation booming in 2025, people with conditions like ADHD, autism, dyslexia and more report a more level playing field in the workplace thanks to generative AI.
A recent study from the UK’s Department for Business and Trade found that neurodiverse workers were 25% more satisfied with AI assistants and were more likely to recommend the tool than neurotypical respondents.
“Standing up and walking around during a meeting means that I’m not taking notes, but now AI can come in and synthesize the entire meeting into a transcript and pick out the top-level themes,” said Tara DeZao, senior director of product marketing at enterprise low-code platform provider Pega. DeZao, who was diagnosed with ADHD as an adult, has combination-type ADHD, which includes both inattentive symptoms (time management and executive function issues) and hyperactive symptoms (increased movement).
“I’ve white-knuckled my way through the business world,” DeZao said. “But these tools help so much.”
AI tools in the workplace run the gamut and can have hyper-specific use cases, but solutions like note takers, schedule assistants and in-house communication support are common. Generative AI happens to be particularly adept at skills like communication, time management and executive functioning, creating a built-in benefit for neurodiverse workers who’ve previously had to find ways to fit in among a work culture not built with them in mind.
Because of the skills that neurodiverse individuals can bring to the workplace — hyperfocus, creativity, empathy and niche expertise, just to name a few — some research suggests that organizations prioritizing inclusivity in this space generate nearly one-fifth higher revenue.
AI ethics and neurodiverse workers
“Investing in ethical guardrails, like those that protect and aid neurodivergent workers, is not just the right thing to do,” said Kristi Boyd, an AI specialist with the SAS data ethics practice. “It’s a smart way to make good on your organization’s AI investments.”
Boyd referred to an SAS study which found that companies investing the most in AI governance and guardrails were 1.6 times more likely to see at least double ROI on their AI investments. But Boyd highlighted three risks that companies should be aware of when implementing AI tools with neurodiverse and other individuals in mind: competing needs, unconscious bias and inappropriate disclosure.
“Different neurodiverse conditions may have conflicting needs,” Boyd said. For example, while people with dyslexia may benefit from document readers, people with bipolar disorder or other mental health neurodivergences may benefit from AI-supported scheduling to make the most of productive periods. “By acknowledging these tensions upfront, organizations can create layered accommodations or offer choice-based frameworks that balance competing needs while promoting equity and inclusion,” she explained.
Regarding AI’s unconscious biases, algorithms can (and have been) unintentionally taught to associate neurodivergence with danger, disease or negativity, as outlined in Duke University research. And even today, neurodiversity can still be met with workplace discrimination, making it important for companies to provide safe ways to use these tools without having to unwillingly publicize any individual worker diagnosis.
‘Like somebody turned on the light’
As businesses take accountability for the impact of AI tools in the workplace, Boyd says it’s important to remember to include diverse voices at all stages, implement regular audits and establish safe ways for employees to anonymously report issues.
The work to make AI deployment more equitable, including for neurodivergent people, is just getting started. The nonprofit Humane Intelligence, which focuses on deploying AI for social good, released in early October its Bias Bounty Challenge, where participants can identify biases with the goal of building “more inclusive communication platforms — especially for users with cognitive differences, sensory sensitivities or alternative communication styles.”
For example, emotion AI (when AI identifies human emotions) can help people with difficulty identifying emotions make sense of their meeting partners on video conferencing platforms like Zoom. Still, this technology requires careful attention to bias by ensuring AI agents recognize diverse communication patterns fairly and accurately, rather than embedding harmful assumptions.
DeZao said her ADHD diagnosis felt like “somebody turned on the light in a very, very dark room.”
“One of the most difficult pieces of our hyper-connected, fast world is that we’re all expected to multitask. With my form of ADHD, it’s almost impossible to multitask,” she said.
DeZao says one of AI’s most helpful features is its ability to receive instructions and do its work while the human employee can remain focused on the task at hand. “If I’m working on something and then a new request comes in over Slack or Teams, it just completely knocks me off my thought process,” she said. “Being able to take that request and then outsource it real quick and have it worked on while I continue to work [on my original task] has been a godsend.”
Over 95% of international data and voice call traffic travels through nearly a million miles of underwater communication cables.
These cables carry government communications, financial transactions, email, video calls and streaming around the world.
The first commercial telecommunication subsea cable was used for telegraphs and was laid across the English Channel between Dover, England and Calais, France in 1850.
The technology then evolved to coaxial cables that carried telephone conversations, and most recently, fiber optics that ferry data and the internet as we know it.
“About ten years ago, we saw the advent of another big category, which is the webscale players and the likes of Meta, Google, Amazon, etc., who represent now probably 50% of the overall market,” said Paul Gabla, chief sales officer at Alcatel Submarine Networks.
Alcatel is the world’s largest subsea cable manufacturer and installer, according to industry trade magazine Submarine Telecoms Forum.
Demand for subsea cables is increasing as tech giants race to develop computation-intensive artificial intelligence models and connect their growing networks of data centers.
Investment into new subsea cable projects is expected to reach around $13 billion between 2025-2027, almost twice the amount that was invested between 2022 and 2024, according to telecommunications data provider firm TeleGeography.
A map of the world’s undersea communication cables.
CNBC | Jason Reginato
Big Tech, big cables
“AI is increasing the need that we have for subsea infrastructure,” said Alex Aime, vice president of network investments at Meta. “Oftentimes when people think about AI, they think about data centers, they think about compute, they think about data. But the reality is, without the connectivity that connects those data centers, what you have are really expensive warehouses.”
In February, the company announced Project Waterworth, a 50,000km (31,000-mile) cable that will connect five continents, making it the world’s longest subsea cable project.
Meta will be the sole owner of Waterworth, which the company says will be a multi-year, multi-billion-dollar project.
Amazon also recently announced its first wholly-owned subsea cable project called Fastnet.
Fastnet will connect Maryland’s eastern shore to County Cork, Ireland, and capacity will exceed 320 terabits per second, which is equivalent to streaming 12.5 million HD movies simultaneously, according to Amazon.
“Subsea is really essential for AWS and for any connectivity internationally across oceans,” Matt Rehder, Amazon Web Services vice president of core networking, told CNBC in an interview about Amazon’s subsea cable investments. “Without subsea you’d have to rely on satellite connectivity, which can work. But satellite has higher latency, higher costs, and you just can’t get enough capacity or throughput to what our customers and the internet in general needs.”
A ship belonging to Alcatel Submarine Networks deploys a plow to install subsea telecommunications cables.
Alcatel Submarine Networks
Google is another large player, having invested in over 30 subsea cables.
One of the company’s latest projects is Sol, which will connect the U.S., Bermuda, the Azores and Spain.
Microsoft has also invested in the infrastructure.
“You’ve seen this huge growth in submarine cables over the past 20 years. And this is driven by just a voracious demand for data,” says Matthew Mooney, director of global issues at cybersecurity firm Recorded Future.
Cut cables
Disruptions due to cable damage can be quite significant, particularly in areas served by few internet connections.
“If you cut a cable, you can cut off multiple countries from internet access, and that includes financial transactions, banking, e-commerce and basic communications,” said Erin Murphy, a senior fellow at the Center for Strategic and International Studies, a nonprofit national security research organization.
That very thing happened to Tonga, an island nation east of Australia.
In 2022, debris from an underwater volcanic eruption severed the island’s only subsea communication cable, cutting the island off from the rest of the world.
In September, cuts to subsea cables in the Red Sea caused disruptions to Microsoft’s Azure cloud service. The company was able to re-route traffic, but users in Asia and the Middle East still faced increased latency problems and degraded performance.
Experts have said that the majority of subsea cable damage is accidental, usually due to fishing activity or a ship accidentally dropping its anchor on a cable. But lately, these cables are becoming the suspected targets of sabotage.
A subsea cable being manufactured at Alcatel Submarine Networks factory in Calais, France.
CNBC
“When you have so many vessels in international waters that are highly trafficked by lots of commercial vessels or fishing vessels, the likelihood of accidents is fairly high,” Murphy said. “But if you’re a hostile actor, you know that as well. So if you’re sending out the so-called Russian ghost fleet, or if you have a Chinese fishing vessel and a cable is accidentally cut, you could just say, ‘Oh, well, it was an accident.’ But it could be intentional. So it’s really hard to discern sometimes when an act of damage is actually intentional or accidental.”
Mooney and Recorded Future have been tracking some of these cases of suspected sabotage.
“I would say that we have seen a significant uptick in what we would consider intentional damages,” Mooney said. “In 2024 and 2025, [we] saw a notable increase in incidents that occurred in the Baltic Sea and around Taiwan. And so it is difficult to be able to determine with 100% validity that these are intentional. However, the fact patterns that emerge from these events does give you cause to be suspicious that they could all be considered accidental.”
Mooney said the increase in suspected sabotage has corresponded to increased tensions between Russia and Ukraine and China and Taiwan.
Despite there being a lack of concrete evidence of subsea cable sabotage, governments are taking the threat seriously.
In January, NATO launched the “Baltic Sentry” following several incidents of cable cuts in the Baltic Sea. The operation involves deploying drones, aircraft and subsea and surface vessels to safeguard the subsea infrastructure in the region.
“As a result, I don’t believe we’ve seen any instances of cable severing since late January 2025, in the Baltic Sea,” Mooney said.
A picture taken on February 4, 2025 shows a Helicopter 15 (HKP15) (L) on the flight deck of patrol ship HMS Carlskrona (P04) on open water near Karlskrona, Sweden, as part of the NATO Baltic Sea patrol mission, the Baltic Sentry, aimed to secure critical underwater infrastructure. The patrol ship HMS Carlskrona (P04) set off from the naval port in Karlskrona on February 4, 2025 to become part of NATO’s Baltic Sentry operation as one of several Swedish ships that are part of Standing NATO Maritime Group One (SNMG1). This is the first time the ship has hoisted the NATO flag on board. The purpose of NATO’s Baltic Sentry operation is to demonstrate presence and secure critical underwater infrastructure. (Photo by Johan NILSSON / TT NEWS AGENCY / AFP) / Sweden OUT (Photo by JOHAN NILSSON/TT NEWS AGENCY/AFP via Getty Images)
Johan Nilsson | Afp | Getty Images
U.S.-China tension
In the United States, the Federal Communications Commission, which is responsible for granting licenses to anyone wishing to install or operate subsea cables connecting to the U.S., has introduced tighter rules on foreign firms building this infrastructure, citing security concerns.
“One area we’ve been particularly focused on are threats that come from the Chinese Communist Party as well as from Russia,” FCC Chair Brendan Carr told CNBC. “So we’re taking actions right now to make it difficult or effectively prohibiting the ability to connect undersea cables directly from the U.S. to a foreign adversary nation.”
Carr said the FCC is also taking steps to make sure the hardware itself isn’t compromised, not allowing Huawei, ZTE or other questionable “spy gear” to be used in undersea cables.
In July, three House Republicans sent a letter to the CEOs of Meta, Amazon, Google and Microsoft asking if the companies have used PRC-affiliated cable maintenance providers.
In response to CNBC’s question about the letter, Meta’s Aime said, “We do not work with any Chinese providers of cable systems on systems that we’ve announced, and we are in full compliance with U.S. policy regulations around partners in the ecosystem and the supply chain.”
Amazon also told CNBC it does not work with Chinese companies.
Microsoft and Google did not return CNBC’s request for comment on the letter.
To understand how subsea cables work, CNBC visited Alcatel Submarine Networks subsea cable manufacturing facilities in Calais, France and Greenwich, England. We also spoke to government officials and tech giants to find out why subsea cables are crucial to keeping us connected and what we can do to protect this critical infrastructure.
Palantir co-founder and CEO Alex Karp attends meetings at the U.S. Capitol in Washington on Oct. 18, 2023.
Jonathan Ernst | Reuters
With Palantir’s stock plummeting more than 11% this week despite a better-than-expected earnings report, CEO Alex Karp took aim at investors betting against the software company.
Karp, who co-founded Palantir in 2003, went after short sellers in two separate interviews on CNBC this week. After “Big Short” investor Michael Burry revealed bets against Palantir and Nvidia, Karp on Tuesday accused short sellers of “market manipulation.”
He repeated that message on Friday in an interview with CNBC’s Sara Eisen, again knocking Burry’s wager against the stock.
“To get out of his position, he had to screw the whole economy by besmirching the best financials ever … that are helping the average person as investors [and] on the battlefield,” Karp said.
Even with Palantir’s slide this week, the stock is up 135% in 2025 and has multiplied 25-fold in the past three years, an extended rally that’s lifted the company’s market cap to over $420 billion. While revenue and profit are growing rapidly, the multiples have shot up much faster, and the stock now trades for about 220 times forward earnings, a ratio that rivals Tesla’s.
Nvidia and Meta, by contrast, have forward price-to-earnings ratios of about 33 and 22, respectively.
In August, Citron Research’s Andrew Left, a noted short seller, called Palantir “detached from fundamentals and analysis” and said shares should be priced at $40. It closed on Friday at $177.93 after late-day gains pushed the stock into the green.
Palantir, which builds analytics tools for large companies and government agencies, reported earnings and revenue on Monday that topped analysts’ estimates and issued a forecast that was also ahead of Wall Street projections.
But the stock fell about 8% after the report and then slid almost 7% on Thursday. Karp told Eisen that the recent boom in Palantir’s share price isn’t just for Wall Street.
“We’re delivering venture results for retail investors,” he said.
While Palantir has in the past faced a fairly heft dose of short interest, there are currently relatively few investors placing big bets against it. The short interest ratio, or the percentage of outstanding shares being sold short, peaked at over 9% in September and is now at a little over 2%, which is about as low as its been since the company went public in 2020.
Still, calling out the doubters is a common occurrence for Karp, who has previously said on CNBC that people should “exit” if they “don’t like the price.”
In May, after the stock plummeted following earnings, Karp said ,”You don’t have to buy our shares.”
“We’re happy,” he said. “We’re going to partner with the world’s best people and we’re going to dominate. You can be along for the ride or you don’t have to be.”
The company has also faced backlash over its work with government agencies like U.S. Immigration and Customs Enforcement, and Karp has admitted that his strong pro-Israel stance led some people to leave the company.
The boisterous CEO has been particularly vocal this week. On Monday’s earnings call, he questioned how happy the people are who didn’t invest in the company, and told them to “get some popcorn.”
And on CNBC he aimed much of his ire at Burry after the investor revealed his short positions in Palantir and Nvidia.
“The two companies he’s shorting are the ones making all the money, which is super weird,” Karp told CNBC’s “Squawk Box” on Tuesday. “The idea that chips and ontology is what you want to short is bats— crazy.”
