Starbucks is accelerating its push into AI-powered back-of-house technology with the rollout of a new automated inventory counting system. The tool, already live in thousands of company-operated locations, will be deployed across all Starbucks-operated stores in North America by the end of September, according to the company.

The system, developed in partnership with Seattle-based NomadGo,  uses handheld tablets combined with computer vision, 3D spatial intelligence and augmented reality to scan shelves and instantly identify what’s in stock. Low-stock items are flagged automatically, helping baristas stay ahead of shortages. Store partners say inventory counts that once took about an hour can now be completed in minutes—often 10 to 15—enabling them to count inventory as much as eight times more frequently than before. Starbucks executives say the technology will eventually be able to automate restock orders as well.

“At Starbucks, technology isn’t about bells and whistles — it’s about creating meaningful efficiencies that preserve and strengthen craft and connection,” Deb Hall Lefevre, the company’s chief technology officer, said in a statement. “With automated counting, we’re delivering both a better experience for partners, and a more consistent, delightful experience for every customer who walks through our doors.”

The implications go beyond speed. More frequent and more accurate counts provide Starbucks with near real-time data pulses that can improve supply chain responsiveness, speed up deliveries and minimize stockouts. Lefevre said the modernization sets the stage for smarter replenishment strategies, ensuring stores remain stocked and staff stay focused on customers rather than clipboards.

The inventory system marks Starbucks’ third major technology announcement in recent months. In May, the company introduced a Shift Marketplace tool that lets baristas swap and manage shifts more easily. In June, Starbucks unveiled Green Dot Assist, a generative AI-powered assistant designed to support store operations, along with a new Next Gen POS system now in pilot.

For Starbucks, the rollout of NomadGo’s technology signals a deeper commitment to using AI and AR to streamline labor-intensive tasks. Automating back-of-house functions has the potential to unlock measurable efficiency gains at scale, especially for a brand that operates nearly 17,000 stores in North America alone.

Competitors are pursuing similar strategies. McDonald’s, for example, is deploying its Restaurant Platform Edge, a Google-powered cloud platform now live in hundreds of U.S. restaurants and expanding globally. One widely publicized feature is its AI-powered “Accuracy Scales,” which weigh outgoing orders and flag discrepancies before they reach the customer. The system is already in thousands of restaurants and has been credited with boosting order accuracy and customer trust. McDonald’s has also piloted a geofencing-based “Ready on Arrival” feature that can cut wait times by more than half.

Chick-fil-A, meanwhile, has been experimenting with computer vision to monitor food quality, as well as geofenced mobile ordering tools to better time kitchen prep and minimize wait times. The chain has also invested in automation around its drive-thru, which remains one of the busiest in the industry.

Together, these initiatives point to a clear trend: the largest quick-service brands are betting heavily on AI, computer vision and AR not just to improve customer-facing speed and personalization, but to streamline the operational backbone of their businesses. For Starbucks, a company that positions its partners and craft at the center of its brand identity, the goal is as much about empowering employees as it is about optimizing efficiency.

“Behind the scenes, these fast, accurate and frequent data pulses automate portions of our operations to speed up deliveries and minimize stockouts,” Lefevre said. “This modernization sets the stage for smarter supply chain optimization and more frequent inventory replenishment — all designed to keep our coffeehouses ready and partners empowered.”

The success of Starbucks’ latest AI deployment will likely be measured not just in saved labor hours, but in fewer stockouts, more consistent guest experiences and the ability to reinvest staff time into service rather than manual tasks. For an industry where a missed latte ingredient or a delayed order can erode trust, the combination of AI, AR and real-time intelligence could prove to be a competitive differentiator in the months ahead.

Nothing we aren’t used to seeing in tech, but still…to apply an old sports idiom to the rapidly evolving world of AI, the best defense is a good offense, no?

Today’s tech news below. —Andrew Nusca

Want to send thoughts or suggestions to Fortune Tech? Drop a line here.

Figma gets crushed in its post-IPO earnings debut

Shares of design software company Figma plunged 14% in extended trading yesterday as investors took a dim view of Figma’s first-quarter earnings report. 

In its fiscal second quarter, Figma’s revenue grew a healthy 41% year-over-year to $249.6 million, roughly in line with analyst expectations. Figma reported $28.2 million in net income, or break-even on a per share basis.  

“We’re at the very start of what I hope is a long-term relationship together,” CEO Dylan Field confidently told listeners as he kicked off the earnings call.

Field, who cofounded the company in 2012 and watched its $20 billion acquisition by Adobe fall apart in 2023, clearly isn’t one to get caught up in the negative. 

“No one knows whether we’re going to look back in five years at everything that’s happening right now in AI and say, ‘Oh my God, those were the bubbliest of times,” Field told Fortune yesterday ahead of the call. “Or: ‘Wow, we totally underestimated the effect it would have on society.’”

Field believes one of the key intersections between AI and design is that AI tools will help broaden access, letting more people become designers. Figma added four new AI-native tools to its platform this quarter and told investors on the call to expect significant investments in AI going forward.

“Our philosophy is that as the models get better, we get better,” he said. “That’s always the test I have strategically for us.” —Allie Garfinkle

Meet the $100m AI startup that wants to kill the billable hour

Eudia, a Palo Alto-based AI startup, is offering something entirely new: the world’s first AI-augmented law firm. 

Its end goal is nothing less than the death of the billable hour that, according to CEO Omar Haroun, has run entirely out of control. 

“Most legal departments have lost control of their budgets and their knowledge,” Haroun said in a press release announcing the launch of Eudia Counsel, which he called “the first AI-native law firm.”

The company has fought hard to bring its novel approach to light, Haroun told Fortune at the company’s 2025 Augmented Intelligence Summit in New York.

Arizona is the only state in the country where a law firm is not required to be owned by lawyers, he said. Even still, there are technicalities. Eudia is not technically set up as a law firm, but a company that is a “provider of a law firm.”

Haroun told Fortune that the economics of AI can, for example, transform pro bono work, which he sees as “the reason people like me went to law school” in the first place.

Gary Hood, general counsel for Berkshire Hathaway-owned Duracell, said using Eudia has been a “no-brainer” for contracts and due diligence during M&A.

Haroun said some clients were spending hundreds of millions of dollars on outside counsel, and that’s where Eudia steps in.

And what about people? Eudia co-founder Ashish Agrawal likened the tools to a brand new employee that every company has to be patient with and incorporate “organically.” Human inputs, he told Fortune, are essential to AI working properly. —Nick Lichtenberg

Is an ‘AI winter’ coming?

As summer fades into fall, many in the tech world are worried about an AI winter. 

There’s a reason this phrase comes so naturally lately: We’ve already lived through several spells of waning enthusiasm and investment in AI over its 70-year history.

The most recent talk has been triggered by growing concerns among investors that AI technology may not live up to the hype surrounding it—and that the valuations of many AI-related companies are far too high. 

Is this chill in the air a passing breeze or the first hints of an impending Ice Age? A look at past AI winters may help. 

For example, there are clear parallels between the hype generated by today’s prominent AI figures and the competing camps working on the technology in the early days of the Cold War. 

There are also historical parallels for recent studies suggesting AI isn’t meeting expectations. In 1966, a committee commissioned by the National Research Council issued a damning report concluding that computer-based translation was more expensive, slower, and less accurate than human translation. 

You can guess what happened to research funding after that.

There are some key differences between then and now. Most significantly, today’s AI boom is not dependent on public funding—though government entities are becoming important customers.

And unlike in the very first AI winter, when such systems were mostly just research experiments, today’s AI is being widely deployed in businesses and homes.

Three years after ChatGPT’s debut, there are certainly a few autumnal signs here and there. But only time will tell if it is the prelude to a deep freeze in AI investment or a momentary cold-snap before the sun appears again. —Jeremy Kahn

More tech

—Apple AI search tool in development. “World Knowledge Answers” reportedly arrives in the spring.

—Instagram for iPadOS arrives. It only took 15 years, Meta!

—xAI CFO Mike Liberatore departs. He joined Elon Musk’s AI startup in April and reportedly exited in late July.

—Scale AI sues former employee for allegedly stealing trade secrets and sharing them with Mercor.

—Tom Siebel steps down at C3 AI, to be replaced by Stephen Ehikian. The company reported a Q1 revenue drop of 19% from the same period a year ago.

—Streameast shuts down. The world’s largest illegal sports streaming platform is knocked out by Egyptian law enforcement and others.

—Mistral valued at €12B. The French AI firm (of Le Chat fame) is reportedly putting the finishing touches on a €2B investment.

Endstop triggered

The global AI chip market is a battleground of geopolitical strategy and technological innovation, with China’s demand for advanced semiconductors emerging as a critical focal point. For investors, the tension between U.S. export restrictions and China’s push for self-reliance creates a paradox: while geopolitical risks threaten to fragment markets, the indispensable nature of cutting-edge AI hardware ensures sustained demand. Nvidia, a leader in AI chip development, finds itself at the center of this dynamic, balancing compliance with its ambition to retain a foothold in China’s $50 billion AI opportunity [2].

Geopolitical Risk: The U.S. Export Control Conundrum

U.S. export restrictions have reshaped the AI chip landscape in China. In Q2 2025, Nvidia reported zero sales of its H20 AI chips to the region, a direct consequence of stringent export controls and the absence of finalized regulatory guidelines for its new licensing agreement [2]. This vacuum has allowed domestic competitors like Cambricon to surge, with the company’s revenue jumping 4,300% in the first half of 2025 [1]. The U.S. government’s 100% tariffs and revocation of VEU licenses have further fragmented global supply chains, compelling firms like AMD and Nvidia to develop lower-performance chips for China while TSMC shifts capital expenditures to the U.S. and Europe [1].

Yet, these restrictions have not eradicated demand for advanced AI hardware. China’s AI industry, supported by state-led investment funds and subsidized compute resources, is projected to grow into a $3–4 trillion infrastructure boom by 2030 [3]. The National Integrated Computing Network, a state-backed initiative, underscores Beijing’s commitment to building a self-sufficient ecosystem [4]. However, bottlenecks persist: limited access to EUV lithography and global supply chain integration remain significant hurdles [4].

Technological Indispensability: The Unmet Need for Performance

Despite China’s strides in self-reliance, the gap between domestic and U.S. semiconductor capabilities remains stark. Companies like Huawei and SMIC are closing this gap—Huawei’s CloudMatrix 384 and SMIC’s 7nm production expansion are notable advancements [1]. However, the performance of these chips still lags behind Nvidia’s Blackwell GPU, which offers unparalleled efficiency for large-scale AI training. This technological disparity has driven Chinese firms like Alibaba to invest in homegrown solutions, including a new AI chip, while still relying on U.S. technology for critical applications [2].

Nvidia’s recent development of the B30 chip—a China-compliant variant of the Blackwell GPU—exemplifies its strategy to navigate these challenges. By adhering to U.S. export restrictions while retaining performance, the B30 aims to secure market access in a landscape where even restricted chips are indispensable [3]. This approach mirrors the broader trend of “compliance-driven innovation,” where firms adapt to geopolitical constraints without sacrificing technological relevance.

Strategic Implications for Investors

For investors, the key lies in assessing how companies balance compliance with innovation. Nvidia’s ability to pivot to the B30 chip highlights its resilience, but the absence of H20 sales in Q2 2025 underscores the fragility of its China strategy [2]. Meanwhile, domestic players like Cambricon and SMIC offer high-growth potential but face long-term challenges in overcoming U.S. export controls and achieving parity with Western rivals [1].

The AI infrastructure boom, however, presents a universal opportunity. As global demand for advanced compute surges, firms that can navigate geopolitical risks—whether through compliance, localization, or hybrid strategies—will dominate. China’s push for self-reliance, while reducing its dependence on U.S. chips, also creates a fertile ground for innovation, with startups like DeepSeek optimizing FP8 formats for local hardware [1].

Conclusion

Nvidia’s experience in China encapsulates the dual forces shaping the AI chip sector: geopolitical risk and technological indispensability. While U.S. export controls have disrupted its access to the Chinese market, the company’s strategic adaptations—such as the B30 chip—demonstrate its commitment to maintaining relevance. For investors, the lesson is clear: the AI race is not just about hardware but about navigating a complex web of policy, innovation, and market dynamics. As China’s self-reliance drive accelerates, the winners will be those who can bridge the gap between compliance and cutting-edge performance.

Source:[1] China’s AI Chip Revolution: The Strategic Imperative and Investment Opportunities in Domestic Semiconductor Leaders [https://www.ainvest.com/news/china-ai-chip-revolution-strategic-imperative-investment-opportunities-domestic-semiconductor-leaders-2508/][2] Alibaba reportedly developing new AI chip as China’s Xi rejects AI’s ‘Cold War mentality’ [https://ca.news.yahoo.com/alibaba-reportedly-developing-ai-chip-123905455.html][3] Navigating Geopolitical Risk in the AI Chip Sector: Nvidia Remains a Strategic Buy Amid Chinese Restrictions [https://www.ainvest.com/news/navigating-geopolitical-risk-ai-chip-sector-nvidia-remains-strategic-buy-chinese-restrictions-2508/][4] Full Stack: China’s Evolving Industrial Policy for AI [https://www.rand.org/pubs/perspectives/PEA4012-1.html]