Capital · United States

AI could upend food delivery

Credited to The Verge · theverge.com · 43m ago

Photograph · The Verge

Delivery apps like DoorDash and Uber Eats have changed how diners eat and how restaurants operate. AI could force a new kind of reckoning for the industry.

DoorDash, the leading food delivery app, processed 970 million orders in its second quarter this year and generated $4.5 billion in revenue. A 10-person startup called Bites is a blip in comparison: It has just around 300 restaurants signed up in the Bay Area, where it’s operating as a pre-seed startup. But this summer, Bites caught DoorDash’s attention.

In August, a slew of restaurants in the Bay Area received a strange email from DoorDash, warning businesses that they may be listed without their consent on Bites.

The form letter, copies of which were seen by The Verge, said that DoorDash had heard from “several partners” that were added to the third-party platform without their knowledge. DoorDash warned restaurants that they may not have agreed to Bites’ “terms, pricing structures, or service standards”; that they may not have control over the accuracy of their menu or hours of operation on Bites; and that, depending on the state, “such practices could be illegal.” The email included instructions on how to contact Bites to request removal.

One way to look at the warning is DoorDash proactively alerting its restaurant partners. But DoorDash going on the offense over a relatively small peer hints at a battle taking shape between established apps and newer, leaner startups looking to upend business in the age of AI.

Tech companies have promised that the future of AI is that ordinary people will have “agents” completing tasks: scheduling meetings, looking for deals on socks, sifting through their inbox and replying to emails, and yes, ordering food on their behalf. If this really does come to bear — if billions of people shift to interact with the internet not through search engines, webpages, and apps, but through large language models — those websites and services will lose control of the relationship with consumers as they stop visiting webpages and apps. What happens to Amazon when it can’t show shoppers boosted item listings because they don’t come to the site? What happens when Uber can’t entice riders with a free subscription trial? What happens when a food delivery service can upsell you to throw in a bottle of wine with dinner? The Verge has coined this brewing fight “the DoorDash problem.” Bites is now the literal embodiment of the term.

Bites’ pitch is simple: Companies like DoorDash are the middleman, forcing restaurants to raise prices to cover the commission the platforms take, leaving customers paying more to make up the difference. DoorDash fees for restaurants range anywhere from 15 to 30 percent per delivery order — the more in commission fees a restaurant forks over, the more diners they are visible to on the platform. Bites, on the other hand, touts simplified delivery and service fees and a flat $1 surcharge per order, which it says allows partners to keep item prices lower. (Diners typically pay the $1 fee, but some restaurants will cover it on a customer’s behalf.) Bites has a traditional app, but it bills itself as “AI native”: Diners can place orders from Bites restaurants directly in ChatGPT, which go not to a third-party platform like DoorDash, but directly to the restaurant.

Bites founder and CEO Bala Subramaniam says the company heard from 14 restaurant partners who had received a letter from DoorDash, and a handful of businesses who were not listed on Bites but now — having been made aware of the company — wanted to join. (Bites also acknowledges some restaurants did request to be removed from the platform; the company says they were listed on Bites from old demos.) Others got emails despite not being listed at all. One of the emails went to Jay Jayaraman, who has 13 restaurants listed on Bites and considers himself an early adopter of the platform.

Before being on Bites, Jayaraman says 80 percent of orders came from DoorDash; now, 65 percent come in through Bites, he estimates, with 25 percent from DoorDash and the remaining sliver from Uber Eats and Grubhub. Jayaraman says his margins are higher with Bites orders than with DoorDash orders.

“We are seeing a huge drop or a huge shift away from DoorDash, which I personally welcome because it’s adding to my bottom line,” Jayaraman says.

LLMs becoming the portal to the web threatens app- and website-based businesses in myriad ways. A shift away from visiting websites and using individual apps means those companies can’t do things like show consumers similar products, offer deals and free trials, and, significantly, monetize via ads and boosted listings. (DoorDash doesn’t only make money via subscriptions and commission fees — its ads business hit $1 billion in 2024.) The most prominent instance of the LLM-service provider tension materializing is the ongoing legal fight between AI company Perplexity and Amazon, in which Amazon sued Perplexity in 2025 to block agents from browsing and shopping on its site. Amazon has also recently blocked Meta’s Muse AI agent from shopping on its site and removed product details from order emails — data that is potentially very valuable for Google’s own agentic shopping features. Amazon’s ad business generated $19.8 billion in revenue in the second quarter of this year, driven by sponsored product listings.

“All of that investment does not have to be made anymore, and yet the pricing has stayed the same.”

Bites can keep prices lower for customers and restaurants because the pricing structure on bigger platforms is outdated, says Matt Maloney, who cofounded both Bites and Grubhub. (Maloney left Grubhub in 2021 after the company was acquired. His involvement with Bites hasn’t been previously reported.) Delivery services spent billions of dollars to build online ordering infrastructure, as well as courting restaurants, diners, and drivers onto their platforms — restaurants raising item prices on delivery apps to offset the high commissions and fees are a relic of that era, Maloney argues. But post-covid lockdown, nobody needs to be convinced of the convenience of delivery.

“All of that investment does not have to be made anymore, and yet the pricing has stayed the same,” Maloney says. Bites partners with point-of-sale systems — companies like Toast and Square that process orders — to pull digitized menus into LLMs and send orders directly to restaurants, instead of through an intermediary like DoorDash. Once a point-of-sale company is partnered with Bites, restaurants click a button to opt in. The Verge tested the service by ordering a pizza and beverage for delivery through ChatGPT using the Bites plugin, which came to a total of $56.73 including tip, tax, and delivery fees. The same order on DoorDash totaled $70.29, which had higher service fees and the same pizza priced $11 more.

“People think it’s the delivery apps that are egregious, [but] it’s actually the model that has made the restaurants actually increase the prices,” Subramaniam, who previously worked at Instacart, says. “And they are actually hungry for a new model.”

Subramaniam says Bites is especially geared toward the diners who already know where they want to eat and don’t need a third-party platform like DoorDash to find them.

“It may not become the discovery platform,” he says. “We don’t need to have sponsored ads … but if you know where you want to eat, we will get you to the place that you want to eat, with the best pricing.”

The explosion of food delivery during the pandemic helped many restaurants stay afloat, but complaints from owners about the platforms are just as common — the high commission fees, the need to buy ads and boost listings, and the wedge that third-party apps create between the people making the food and the people eating it. Platforms like DoorDash withhold most data from the restaurants, typically only sharing a customer’s first name, last initial, and what they ordered. That data is valuable to restaurants: A business could use direct contact information to send discounts or notices to customers, develop loyalty programs, or offer customer support if there are problems with the food that’s delivered. Instead, delivery apps hold on to that information. This has spilled into the courtroom, too: A New York City law attempted to compel third-party platforms to share customer information with restaurants upon request, but was struck down by a judge. Jayaraman, the restaurateur, says he gets “absolutely nothing” from DoorDash when it comes to customer data; with Bites he is able to contact customers directly because the orders (and customer contact information) come straight to him.

Platforms like DoorDash withhold most customer data from restaurants

DoorDash didn’t respond to specific questions around its pricing structure, what data it shares with restaurants, or how AI might disrupt the app-based model of ordering takeout. Spokesperson Lisa Singh told The Verge that “some restaurant partners raised concerns” about appearing on Bites without their consent. “The information shared with partners made clear that no action was needed if they had opted in,” Singh said.

DoorDash itself has signaled agentic ordering is a priority: The company announced in September it was working on a way for diners to place orders within iMessage using prompts like “reorder my usual from [restaurant].” Even this upends the business model — if an agent does the ordering for diners who already know what they want, what does a restaurant get for spending money boosting its business on the DoorDash app? A major benefit that delivery platforms provide is discovery. It’s still not clear what that looks like in an agentic future.

For Jayaraman, there is no love lost for the current business model of delivery apps — but even as the portion of takeout that comes through DoorDash dwindles, he plans to stay on the platform to serve those customers. The AI industry has made lofty promises to the public — automating chores and tasks, saving time and money while shopping, and providing endlessly compliant, personalized virtual assistants at humans’ beck and call. Whether that becomes a reality depends on both customers and goods and service providers agreeing that it’s the future. Jayaraman has not yet turned on ChatGPT ordering, saying he wants to see a live demo from Bites first and hear its pitch.

Subramaniam and Maloney see the emails sent by DoorDash as proof that they’re on the right track. Jayaraman, too, took the email as a sign that the market leader was worried.

“I didn’t respond,” Jayaraman says of receiving the email from DoorDash. “I was very happy when I received the letter, because DoorDash is feeling the pain.”

CapitalUnited States

Vinci reels in $250M for its engineering simulation platform

Credited to SiliconANGLE · siliconangle.com14h ago

Useful

Vinci4D Inc. today announced that it has closed a $250 million funding round at a $1.5 billion valuation.

Advent, Temasek and Xora jointly led the Series B deal. They were joined by several startup funds including AMD Ventures.

Hardware engineers use simulations to find issues in their product designs. For example, a car designer might simulate fog to evaluate the impact of adverse weather on a vehicle sensor. Historically, setting up simulations required highly specialized skills and a significant amount of time. As a result, many hardware teams only use the technology intermittently.

Palo Alto, California-based Vinci is looking to change that. Its namesake software platform can run simulations after every major change to a product blueprint. If engineers run a simulation the day after a design change rather than two weeks later, they can detect issues that much faster. That lowers product development times.

Vinci’s initial focus is speeding up chip projects. In particular, the platform helps engineers simulate the thermal properties of processors, memory modules and interconnects. A chip team could, for example, study whether a graphics card might overheat when it operates near its peak frequency.

Vinci is powered by a physics-focused artificial intelligence model. Developers can upload a CAD drawing of a chip, input certain related data and have the AI automatically turn it into a simulation.

One of the most time-consuming tasks involved in simulation projects is meshing. That’s the process of breaking up a product blueprint into numerous small, three-dimensional shapes. That step is necessary because a large array of small shapes can be processed much faster than a continuous surface such as a chip.

Vinci performs meshing automatically. From there, it uses custom graphics card kernels to speed up convergence, which is the process of finalizing simulation calculations. Kernels are the low-level software modules that make up an AI model.

Original · SiliconANGLE

Useful

Vinci4D Inc. today announced that it has closed a $250 million funding round at a $1.5 billion valuation. Advent, Temasek and Xora jointly led the Series B deal. They were joined by several startup funds including AMD Ventures. Hardware engineers use simulations to find issues…

View the articleSiliconANGLE

Front Method

Power & Sites · United States

LG Electronics Signs 5GW Cooling Deal for North American AI Data Centers

Credited to TheElec · thelec.net · 15h ago

LG Electronics plans to pursue a cooling-solutions supply business totaling 5 gigawatts (GW) in North America.

LG Electronics said Oct. 6 that it had signed a definitive long-term agreement to supply chillers for AI data centers (AIDC) with infrastructure platform company Air Control Concepts (AIR).

Under the agreement, LG Electronics will gradually supply cooling solutions, including chillers used to manage heat in data centers, to facilities in the US and Canada. The company did not disclose specific delivery schedules, customers or contract values.

Air Control Concepts is an infrastructure platform company that oversees the full lifecycle of AIDC projects in North America, including design, construction and public procurement. It carries out large-scale projects and counts major AIDC companies around the world among its key customers. The company specializes in cooling and HVAC infrastructure solutions.

LG Electronics began developing its chiller business in earnest in 2011. It started supplying cooling solutions this year for a major North American AIDC project secured last year. The company also supplies cooling solutions to markets in the Global South, including the Middle East and Indonesia.

According to market research firm Omdia, global power capacity required by AIDCs is projected to rise by 108 GW, from 226 GW in 2026 to 334 GW in 2027.