Instead, Amazon will automatically pull those costs from sellers’ sales proceeds. Credit or debit cards will be used only as a backup if seller proceeds fall short. The new card rules, which were originally set to go into effect April 15, have since been pushed back to Aug. 1 after sellers complained. However, the underlying shift to deducting ad costs from proceeds is seen as inevitable, and its implementation could threaten the working capital structure many depend on.

The new rules effectively reroute payment flows from a credit-based system to an account-to-account (A2A) mechanism. For large enterprises with diversified balance sheets, the shift may register as a marginal operational tweak. But for the small and medium-sized businesses (SMBs) that form the backbone of Amazon’s marketplace, the implications are more profound.

At its core, the change removes two financial tools that sellers have come to rely on: float and rewards. Credit cards, often viewed as mundane transactional instruments, serve a more strategic purpose for these firms. They provide a grace period between incurring expenses and settling them, smoothing cash flow and enabling reinvestment into inventory, marketing and operations. Rewards, whether cash back or points, also function as a modest but meaningful rebate on costs.

By bypassing cards entirely, Amazon is effectively reclaiming that value for itself while shifting the burden of liquidity management back onto sellers.

Read more: Mid-Tier Retailers Caught Between Amazon and Walmart

Unpacking Amazon’s New Card Rules

Amazon’s decision underscores a larger shift in how marketplaces are positioning themselves. Increasingly, they are not just platforms for transactions but orchestrators of financial flows. By controlling how money moves within their ecosystems, they gain leverage over participants’ liquidity, risk exposure and ultimately, their viability.

Amazon’s new card rules serve as a reminder that the financial underpinnings of marketplace participation are as important as the mechanics of selling. They highlight the need for businesses to diversify their financial tools and remain vigilant about changes that could impact liquidity.

For sellers, the challenge lies in the asymmetry of power. Participation in a dominant marketplace often requires adherence to its evolving rules, even when those rules introduce new constraints. The result is a form of financial dependency in which the platform’s decisions can reshape the contours of a business overnight.

Crucially, the ad billing change didn’t happen in isolation, and came on the heels of Amazon’s new DD+7 (delivery date plus seven days) payouts that took effect in March, and a new 3.5% fuel and logistics surcharge that is reportedly being implemented due to the impact of the ongoing Middle East conflict on energy prices.

The episode also reveals a broader truth about the evolving role of credit cards in business operations. Far from being mere payment tools, cards have become a form of strategic infrastructure. They enable businesses to manage cash flow, access short-term financing, and optimize expenses in ways that are deeply integrated into their operational models.

Among the key takeaways from “SMB Growth Monitor: Small Businesses, Big Credit Needs,” a PYMNTS Intelligence collaboration with i2c, was the fact that 53% of business credit card use is preplanned, suggesting that  SMB card spending is often deliberate and tied to business operations rather than impulse purchases.

See also: Amazon and Walmart Push AI Deeper Into the Shopping Aisle 

Marketplaces Become Financial Gatekeepers

At a more technical level, Amazon’s shift can be viewed through the lens of interchange fees. When a transaction is processed via a credit card, the merchant, or in this case, Amazon as the billing entity, pays a fee to the card issuer and network. These fees, typically a few percentage points of the transaction value, represent a significant cost at scale.

By moving to an A2A model, Amazon sidesteps these fees entirely. Payments are effectively internal transfers within its ecosystem, executed without the involvement of card networks. The savings are substantial, particularly given the volume of advertising spend flowing through the platform.

From Amazon’s perspective, the logic is straightforward: reduce transaction costs, improve margins and streamline payment flows. From the seller’s perspective, however, the calculus is less favorable. The cost savings achieved by Amazon come at the expense of the financial flexibility that cards provide.

Looking ahead, Amazon’s move may foreshadow a broader transition toward A2A payments as a foundation for “agentic commerce,” an emerging future in which automated systems handle transactions on behalf of businesses and consumers. In such a world, efficiency and cost minimization become paramount, and traditional intermediaries like card networks may play a reduced role.

For small and medium-sized sellers, the stakes are high. Access to working capital is a critical determinant of growth and survival. When that access is constrained, even temporarily, the effects can cascade through inventory management, marketing and customer fulfillment.

The Growth Corporates Working Capital Index by PYMNTS Intelligence and Visa shows that 85% of middle market firms are using working capital solutions.

Whether this model becomes the norm remains to be seen. What is clear is that payment mechanisms are no longer a peripheral concern; they are central to the dynamics of modern commerce.  For now, the lesson is both simple and complex: in the world of digital marketplaces, the rules of payment are the rules of power.

Amazon outlined this and other trends among sellers in a press release outlining findings from its latest annual Small Business Empowerment Report.

“Amazon has long invested in helping sellers grow and thrive,” Mary Beth Westmoreland, vice president, worldwide selling partner experience at Amazon, said in the release. “Now, generative AI is reshaping what’s possible for small businesses, putting powerful tools in their hands that were once reserved for larger retailers.”

Amazon’s Seller Assistant, an agentic AI expert for sellers that can do things like provide sales projections or offer advice on product improvements, had more than 230,000 monthly users in 2025. When the AI assistant recommended actions, sellers followed its advice 90% of the time.

The company’s generative AI listing tools, which create high-quality listings in a fraction of the time it would take sellers to do it manually, were used by independent sellers to create more than 12 million listings in 2025.

Other AI-powered tools highlighted in the release include a Product Opportunity Explorer that surfaces unmet demand, emerging trends and pricing insights based on its analysis of billions of customer interactions; an onboarding assistant that helps new sellers by answering questions, explaining requirements and guiding them through the process; and Packaging Recommendations that save sellers money by optimizing how products are packed and shipped.

Independent sellers account for more than 60% of the sales in Amazon’s store, according to the release.

During 2025, the number of sellers with over $1 million in sales on the platform grew by 36% to surpass 75,000, the average annual sales of all independent sellers in the U.S. increased by 30% to top $375,000, and more than 11,000 U.S.-based sellers gained a tenfold increase in sales.

These sellers supported more than 2 million jobs across the country, per the release.

Amazon CEO Andy Jassy said in a 2025 Letter to Shareholders posted April 9 that AI will play a role in all the company’s businesses, including retail, cloud computing, advertising and autonomous ride-hailing.

“It will reshape every customer experience we offer and unlock entirely new ones,” Jassy wrote.

On Wednesday (April 29), during Amazon’s earnings call, Jassy highlighted the company’s new Seller Central experience and said it is the kind of tool that turns AI into a seller-service product.

But beneath the talk of Trainium, Bedrock and data centers was a more revealing retail story: Amazon, like everyone else, is still trying to figure out what agentic AI will actually mean for shopping, advertising and the future of eCommerce.

CEO Andy Jassy made clear that Amazon sees AI agents as the next major interface between consumers and commerce. He called the opportunity in agentic commerce “very good for customers,” and said he expects it to be good for Amazon as well. The clearest expression of that strategy is Rufus, Amazon’s AI shopping assistant, which Jassy said has improved substantially over the past year. Monthly active users for Rufus are up more than 115%, while engagement is up nearly 400% year over year.

The bigger issue is where shoppers will start when they want an AI agent to help them buy something. Jassy acknowledged that Amazon is talking with third-party horizontal agents, but he was blunt about their current limits. He compared the moment to the early days of search-engine referrals to eCommerce. Search got better over time, he said. Agentic commerce is not there yet.

Third-party agents, Jassy said, still have trouble getting “the pricing right or the product information right.” They also lack the personalization data and shopping history that Amazon has built over decades. That gap may be the core of Amazon’s strategy. The company does not need to win agentic commerce from scratch. It needs to turn its existing customer relationship into a better shopping agent than a general-purpose tool can offer.

Data Makes Rufus Run

That is Amazon’s opening. Jassy argued that shoppers may prefer to begin with an agent tied to a retailer they already use because that retailer has better data. Amazon knows what customers bought, what they viewed, what similar customers purchased, where they ship items and how their accounts are configured. That gives Rufus a practical advantage that a general-purpose agent may not have.

Jassy put the ambition plainly: Amazon wants Rufus to become “the best shopping assistant anywhere.” The comment captured Amazon’s broader view of agentic commerce. The company is not treating it as a separate channel. It is treating it as a new layer on top of the shopping experience it already controls.

The earnings materials showed how far Amazon is already pushing AI into the store. The company said Rufus can research products, track prices and automatically buy items in Amazon’s store when they reach a set price. Amazon also introduced sponsored products and brand prompts inside Rufus. Nearly 20% of shoppers who interact with a brand prompt in Rufus continue the conversation about that brand, according to the company.

That opens a different kind of advertising opportunity. In traditional search, a shopper types a query, scans a results page and clicks. In agentic shopping, the process may unfold through a conversation. Jassy said these experiences tend to be “multi-turn conversations,” with the customer narrowing the need as the agent asks and answers follow-up prompts. That creates more chances to surface relevant products, including sponsored recommendations.

The seller side is part of the same strategy. Amazon introduced a Seller Central experience that dynamically generates personalized visualizations of data, insights and scenarios based on a seller’s goals. Jassy said the early response and feedback have been strong. For Amazon, that kind of tool turns AI into a seller-service product. For sellers, it could make Amazon’s marketplace feel less like a dashboard and more like an operating system.

AWS Drives the Strategy

The broader AI strategy, however, still runs through AWS. Jassy said Amazon has “never seen a technology grow as rapidly as AI,” and he framed agents as one of the biggest reasons customers are choosing AWS. Bedrock, Amazon’s platform for building with AI models, saw customer spending grow 170% quarter over quarter and processed more tokens in the first quarter than in all prior years combined, according to the company.

Jassy also described a shift from simple AI responses to systems that remember context and take action. In the call, he said the future of agents is stateful, meaning they can store identity, remember prior interactions and use tools to complete tasks. That is central to Amazon’s pitch for Bedrock Managed Agents, which the company announced in preview. It is also central to why Amazon sees AI as a long-term driver of both cloud spending and commerce activity.

That is the connective tissue in Amazon’s earnings call. AWS is selling the infrastructure. Rufus is becoming the shopping interface. Seller Central is becoming more automated. Amazon Ads is preparing for conversational discovery. None of it suggests agentic commerce is fully formed. But it does show that Amazon is trying to own the parts of the stack that could matter most: data, compute, shopping intent, seller tools and advertising.

Three other parts of the call showed how Amazon is using its core operations to support that shift:

• Delivery speed: Jassy said Amazon has delivered more than 1 billion items same day or overnight so far this year. The company also announced 1-hour and 3-hour delivery options on more than 90,000 items in the U.S., with 1-hour delivery available in hundreds of cities and towns and 3-hour delivery in more than 2,000.
• Alexa: Alexa+ expanded to Mexico, the U.K., Italy and Spain. Jassy said customers using Alexa+ are talking to Alexa twice as much, completing purchases on devices three times more, streaming music 25% more and using smart-home functions 50% more than with Alexa.
• Grocery: Amazon said it generated more than $150 billion in gross grocery sales in 2025, making it the second-largest grocer in the U.S. Jassy said perishable sales have grown more than 40 times year over year where same-day perishables are available, and those shoppers add nearly three times as many items to their orders and spend more than 80% more than customers who do not buy same-day perishables.

The financial results gave Amazon plenty of room to keep investing. Net sales rose 17% to $181.5 billion in the first quarter, while operating income increased to $23.9 billion from $18.4 billion a year earlier. AWS sales grew 28% to $37.6 billion, and AWS operating income reached $14.2 billion. Net income rose to $30.3 billion, or $2.78 per diluted share, including $16.8 billion in pre-tax gains tied to Anthropic investments. Free cash flow fell sharply to $1.2 billion for the trailing 12 months, driven largely by higher property and equipment spending tied to AI.

This feature already provided AI-generated audio summaries of key product features. Now, with the addition of a “Join the Chat” feature, it can also answer questions during those summaries, Amazon said in a Tuesday (April 28) blog post.

When customers ask questions during the summary, the AI host will pause, answer the question in real time, and then pick up where it left off with the summary.

Join the Chat is now available to U.S. customers in the Amazon Shopping app, on both iOS and Android devices. Hear the Highlights, of which Join the Chat is a part, is available for millions of products, though not all products.

To use these tools, app users go to a product detail page; tap play on the Hear the Highlights button, if there is one, to hear the audio summary; tap the raised-hand icon to initiate Join the Chat; and either type or use the microphone icon to ask a question.

Rajiv Mehta, vice president, conversational shopping at Amazon, said in the post that customers can use Join the Chat at any point during a Hear the Highlights audio summary to ask questions such as “Do people find this sweater itchy?” or “Is this product dishwasher safe?”

“The AI hosts incorporate the question into the conversation, deliver a tailored response grounded in product details, customer reviews and other publicly available information, and then seamlessly continue the episode,” Mehta said. “Customers aren’t interrupting the experience — they’re part of it.”

Amazon began testing the Hear the Highlights feature in May 2025, saying these AI-powered voice summaries make “product research fun and convenient — it’s like having helpful friends discuss potential purchases to make your shopping easier.”

The company extended Hear the Highlights to all customers in May 2025, with some observers saying this feature could replace user-created reviews as the main source of product information.

The PYMNTS Intelligence report “Getting to Know You: How AI Is Shaping the Future of Shopping” found that voice commerce is one of the ways in which retailers are responding to consumers’ demand for personalized, AI-driven experiences that offer convenience, customization and efficiency.

The new artificial intelligence (AI) tools, announced by the company’s Amazon Web Services arm Tuesday (April 28), are aimed at logistics workers, human resources, customer service departments, along with a previously announced healthcare offering.

“These new Connect solutions draw on our expertise incorporating agents throughout Amazon’s operations,” the company said, noting the breadth of its experience in supply chain, hiring, custom interactions and its health and pharmacy business.

“We’ve built AI systems not just to think about these challenges but to solve them in the real world, at scale, every single day.”

A central component of this expansion is Amazon Connect Decisions, a tool designed to manage supply chain logistics and forecasting.

The company notes that supply chain disruptions can take weeks to resolve due to fragmented data systems and manual spreadsheet coordination, while its solution uses more than 25 specialized supply chain tools and foundation models from Amazon’s Supply Chain Optimization Technologies (SCOT) to automate these processes

Amazon Connect Talent is modeled after the company’s internal hiring processes, the release added, citing the company’s annual recruitment of hundreds of thousands of workers for the holiday shopping season.

Amazon Connect Customer — previously known simply as Amazon Connect — has been updated to include “new configuration capabilities that enable organizations to set up conversational AI in weeks, not months, and configure experiences without technical expertise.”

Amazon Connect Health was introduced in March to automate administrative tasks for healthcare providers. As covered here, this tool handles patient verification, scheduling, medical histories, documentation and coding.

The new offerings come as AI agents play an increasingly prominent role in the workplace, as PYMNTS wrote earlier this week.

Agentic AI “is starting to move from conference-room promise to operating-room reality in financial services, where banks, insurers and asset managers are testing software agents on the manual work that slows down decisions,” the report said.

The report cited recent findings from the likes of Snowflake, KPMG and The Economist based around the same theme: the first major gains are likely to be the result of assigning agents tightly controlled tasks like gathering data, checking documents, routing approvals and preparing recommendations.

“The larger shift is not simply faster automation,” PYMNTS wrote. “It is a new model for financial work, one in which firms use stronger data foundations, clearer governance and human oversight to turn fragmented processes into more continuous workflows.”

Amazon built one system to handle all of it.

Through its One Medical arm, the company launched a GLP-1 Management Program bundling clinical screening, prescriptions, ongoing supervision and pharmacy delivery into a single offering. The program connects GLP-1 treatment to patients’ broader primary care, proactively managing how weight loss intersects with cardiovascular health, metabolic conditions and overall health. Amazon isn’t just filling prescriptions. It’s owning the care journey.

What Amazon Built

The program runs end-to-end. Patients get pre-visit screening, structured consultations, regular follow-up visits and evidence-based treatment protocols with built-in safety checks. Individuals who already hold a GLP-1 prescription but aren’t Amazon One Medical primary care patients can access on-demand prescription renewals through telehealth, starting at $29 for message consultations or $49 for video visits, according to Amazon.

The pharmacy layer runs parallel. Amazon Pharmacy offers oral GLP-1 medications including Wegovy and Foundayo with insurance coverage starting at $25 per month or cash-pay options starting at $149 per month. Injectable options including Wegovy and Zepbound start at $299 per month with cash-pay pricing.

Fulfillment is the final piece. Same-day delivery is available in nearly 3,000 cities today and will expand to roughly 4,500 by the end of 2026, Amazon said. Customers can compare insurance and cash-pay prices side by side at checkout. The company says its automatic coupon program has saved Amazon Pharmacy customers more than $200 million to date.

Obesity affects more than 40% of U.S. adults and contributes to nearly $173 billion in annual medical costs, Amazon noted in its announcement. GLP-1 therapies address that population but have historically reached patients through fragmented channels, inconsistent pricing and recurring access gaps. Amazon’s model cuts across all three.

Pressure on Drugmaker Channels

The launch landed hard on pharmaceutical stocks. Shares of Eli Lilly and Novo Nordisk declined on the day of the announcement, as Amazon’s integrated model raised questions about the direct-to-consumer strategies both companies have built, according to Invezz.

Both drugmakers have invested in their own patient-facing platforms. Eli Lilly runs LillyDirect. Novo Nordisk operates NovoCare. Those platforms were built to maintain a direct relationship with patients and reduce dependence on intermediaries. Amazon’s model may shift clinical decision-making toward providers inside its own network, reducing the reach of pharmaceutical marketing.

The timing adds pressure. Novo Nordisk is already in market with an oral GLP-1 pill for obesity, hitting 50,000 weekly prescriptions within weeks of launch, while Eli Lilly’s oral option is expected to follow later this year, according to CNBC. The oral format lowers the barrier to entry for new patients. It also fits Amazon’s fulfillment model more cleanly than injectables. A pill ships in an envelope. A KwikPen requires cold chain logistics.

What It Signals for Health Commerce

Amazon has assembled the pieces before. Pharmacy. Primary care through One Medical. Logistics. The GLP-1 program is the first time those pieces run as a single product around a specific chronic condition.

The Trump administration announced pricing agreements with Eli Lilly and Novo Nordisk that will cut GLP-1 prices to as low as $150 per month and expand Medicare coverage through a pilot program starting mid-2026, according to Patient Care Online. Lower prices expand the patient pool. A larger patient pool rewards whoever controls the access point.

Amazon controls the access point. The program is available now at One Medical locations across the U.S.

 

In its first “Trustworthy Shopping Experience” report, Amazon said it is using artificial intelligence (AI), tighter seller checks and wider enforcement efforts to stop many of those threats before they reach shoppers or merchants.

The report marks a shift from Amazon’s earlier brand protection reports, which focused largely on counterfeits and intellectual property. The new version keeps those themes but widens the frame to include what Amazon describes as four connected priorities: proactive controls, tools that anticipate risk, action against fraudsters and consumer education.

Claire O’Donnell, Amazon’s vice president of selling partner trust and store integrity, said that broader framing better reflects how the company already thinks about the problem.

“The approach we’ve taken to the report this year for the first time is this much broader look at our trust and safety efforts,” O’Donnell told PYMNTS. “A key cornerstone … is how do we get more proactive” so Amazon can identify issues with sellers or products ahead of time and keep them from reaching customers.

That push starts before many sellers even begin operating on the platform. Amazon said all new sellers must complete a verification process designed to make it easier for legitimate businesses to join while making it harder for fraudsters to get in. The company said it checks identity, business links and payment flows, then continues monitoring sellers over time.

O’Donnell said the goal is to add protection without burdening honest merchants.

“Our goal is those verifications are seamless for legitimate selling partners, but they’re very difficult for bad actors to game,” she said. “Are you a real person? Are you connected to a real business? Does the money flow to that real business?”

AI Comes to the Rescue

AI is playing a larger role in that effort. Amazon said its systems scan billions of attempted product page changes each day, while multimodal tools analyze text, images, seller behavior and supply chain patterns for signs of abuse. The company also said its Omniscan machine learning system has generated image sets for more than 12 million products to help verify required safety information before listings go live in several markets.

The company is also emphasizing prediction, not just detection. Amazon said an early warning system that pulls in signals from social media and other retailers helped it block infringing listings tied to a viral branded product eight days before the brand owner had even shared its intellectual property with Amazon. The report also pointed to SENTRIX, an AI system that Amazon said has improved its ability to identify and remove phishing sites, helping lift successful takedowns of phishing URLs by more than 10%.

For O’Donnell, the message is that these threats increasingly overlap.

“I think that’s exactly right,” she said when asked whether trust and safety now looks more like one connected problem than a series of separate ones. “We’re just sharing more of that interconnected trust and safety strategy externally.”

Amazon used the report to underscore the scale of its enforcement work. It said its Counterfeit Crimes Unit has pursued more than 32,000 fraudsters through lawsuits and criminal referrals since 2020 across 14 countries. In 2025 alone, the company said it identified and disposed of more than 15 million counterfeit products worldwide. It also said legal action helped shut down more than 100 websites tied to fake reviews and scams targeting the Amazon store, while its systems blocked hundreds of millions of suspected fake reviews before they appeared online.

Connecting With Customers

Amazon is also trying to show that trust extends beyond the transaction itself. The company said it directly contacted millions of customers in 2025 with product safety information and worked with 34 consumer organizations on 71 safety topics across seven countries. Its recalls and safety alerts page is designed to notify affected customers when governments announce recalls and direct them to refund, return or repair options.

Still, O’Donnell said the hardest challenge may be one that no platform can solve alone.

“It’s really organized retail crime,” she said. “It’s a coordinated criminal enterprise.”

She said retailers, brands and law enforcement will need to work together to stop networks that target supply chains, logistics systems and digital storefronts at scale.

That may be the central point of Amazon’s new report. Marketplace fraud is no longer just a counterfeit problem or a seller problem. It is a broader trust problem, and one that large platforms increasingly believe must be tackled before customers ever see the threat.

“The Attorney General’s motion is a transparent attempt to distract from the weakness of its case, coming more than three years after filing its complaint and based on supposedly ‘new’ evidence it has had for years,” an Amazon spokesperson said in a statement emailed to PYMNTS.

Bonta said Monday a press release issued by the California Department of Justice that evidence released publicly and largely unredacted Monday details how Amazon pressures vendors and competitors to increase prices so that they are not lower than its own.

Bonta said the evidence is in a filing that he filed in a request for a preliminary injunction, which asks a court to halt what he called Amazon’s illegal conduct while the lawsuit is underway. On Monday, he secured public access to the filing, according to the release.

“The company is price fixing, colluding with vendors and other retailers to raise costs for Americans beyond what the market requires — beyond what is fair,” Bonta said in the release. “Amid a crisis of affordability, Amazon is illegally working to rake in profits by making sure consumers have nowhere else to turn to for lower prices.”

According to the California Department of Justice press release, Amazon caused price increases for consumers via three illegal schemes, including colluding with competitors through their common vendor to match each other’s increased price, asking a competitor through a vendor to raise their retail price so that Amazon can do the same, and having a vendor remove a product from a competing retailer that is offering a lower price.

The department said in the release that these practices span years and product lines.

In the Monday statement, the Amazon spokesperson said: “Amazon is consistently identified as America’s lowest-priced online retailer, and we’re proud of the low prices customers find when shopping in our store. Amazon looks forward to responding in court at the appropriate time.”

The Guardian reported Thursday (April 16) that Bonta said that documents that had not yet been made publicly available showed that Amazon pressures sellers using its platform to raise their prices on competitors’ websites if those prices are even a penny lower than Amazon’s.

Bonta filed the antitrust lawsuit in September 2022. On Thursday, it was reported that a California court ruled that key factual disputes must be resolved before the case can proceed, which denied Amazon an early victory in the antitrust case.

According to the Monday press release, a hearing on California’s preliminary injunction motion is scheduled for July and the case is set to go to trial in January 2027.

Amazon Bio Discovery, announced last week, is designed to offer scientists direct access to a wide range of specialized AI models known on biological foundation models (bioFMs) that are trained using large biological datasets. 

“These models generate and evaluate potential drug molecules, known as candidates, helping scientists accelerate antibody therapies during the early stages of drug discovery,” the announcement said. “But access alone is not enough.”

Amazon Bio Discovery allows scientists to converse naturally in their preferred terminology with an AI agent to choose the appropriate models for their research goals, optimize inputs and gauge candidates for experimentation, the announcement said.

In addition, scientists also train models on their prior experimental data for more accurate predictions, and send candidates to physical labs for synthesis and testing, “with results routing back to the application for rapid iteration, creating a lab-in-the-loop experimentation cycle,” the announcement added.

“AI agents make powerful scientific capabilities accessible to all drug researchers, not just those with computational expertise,” Rajiv Chopra, vice president of AWS healthcare AI and life sciences, said in the announcement.

“These AI systems can help scientists design drug molecules, coordinate testing, learn from results, and get smarter with each experiment. This combination of cutting-edge AI and the robust, secure infrastructure AWS has built for regulated industries allows scientists to accelerate antibody discovery in ways that weren’t possible before.”

The announcement comes as pharmaceutical companies are reshaping their operating models around AI to offset the cost of drug development, as PYMNTS wrote earlier this year.

“By embedding machine learning into trial execution and compliance infrastructure, drugmakers are targeting the most costly and failure-prone bottlenecks in how therapies are tested, reviewed and ultimately brought to market,” that report said.

Big Tech and hardware players are entering these workflows, blurring the boundaries between IT and life sciences. For instance, Nvidia and Eli Lilly announced a co-innovation lab to drive drug discovery.

Meanwhile, Google’s research division is using Gemma AI models for cancer therapy discovery, showing how large-language and generative models can analyze biological pathways and suggest novel therapeutic hypotheses.

“Taken together, these developments point to a broader reality: AI is no longer a niche computational aid in early R&D,” that report said. “It is becoming an end-to-end operational ecosystem that supports patient selection, safety monitoring, documentation generation, trial logistics and regulatory engagement.”

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.

The documents have not yet been made publicly available, according to the report. They were obtained as part of a civil case filed by California in 2022, alleging that Amazon engages in price fixing, per the report.

According to the report, documents show that one seller testified that Amazon stopped making one of his products a featured item in the retailer’s Buy Box because that product was a cent cheaper on Walmart; another seller said that a product that had been “suppressed” by Amazon was reinstated to the Buy Box after he raised the price on Wayfair; and an Amazon engineer said the company uses Buy Box suppression to discourage sellers from using Temu.

California Attorney General Rob Bonta told The Guardian that these documents reinforce the state’s claims that Amazon “unfairly punishes sellers whose products are sold at lower prices by other online retailers.”

Amazon said in a statement provided to The Guardian that its practices promote competition and that the company works to provide customers with low prices and a quality customer experience.

“Just like any store owner who wouldn’t want to promote a bad deal to their customers, we don’t highlight or promote offers that are not competitively priced,” the company said, per the report. “It’s part of our commitment to featuring low prices to earn and maintain customer trust.”

When Bonta filed the antitrust lawsuit in September 2022, he said that Amazon has policies that prohibit merchants from selling products at lower prices on other online platforms or their personal websites and that there was ample evidence indicating the company’s practices lead to consumers paying more for various products.

On Thursday, it was reported that a California court ruled that key factual disputes must be resolved before the case can proceed, which denied Amazon an early victory in the antitrust case. The court found that there were too many unresolved factual disagreements about whether the company’s policy promotes competition or suppresses it, per the report.

A preliminary injunction is set to be heard in July, and trial is set for January 2027, according to a Thursday press release from the State of California Department of Justice.