SUCCESS

The vast majority of business data is tabular — living in data warehouses, CRMs, and financial ledgers — yet building a reliable model from it still means training a new one from scratch for every dataset, then maintaining hyperparameter tuning loops, feature engineering, and retraining pipelines to fight data drift. Google Research is proposing a way around that: a new foundation model called TabFM that treats tabular prediction as an in-context learning problem instead.It can generate predictions for a new, unseen table in a single forward pass. For enterprise developers and AI engineers, this reduces the time-to-production from weeks of pipeline engineering to a single API call.The challenge with traditional MLTo extract reliable predictions from a gradient-boosted tree, data scientists must build and maintain complex data pipelines. They have to clean messy inputs, impute missing values, encode categorical variables into numerical formats, and engineer custom feature crosses.Once the data is ready, they must run repetitive hyperparameter optimization loops, searching across learning rates, tree depths, subsampling ratios, and regularization grids to find the best configuration. Once deployed, these traditional models “incur ongoing operational debt through data drift monitoring and retraining pipelines to stay accurate,” Weihao Kong, Research Scientist at Google Research, told VentureBeat.Meanwhile, the rest of the AI industry has moved on. Generative AI models for text and computer vision have seamlessly shifted to zero-shot inference, where a model can perform a completely new task simply by being prompted with context. Large language models (LLMs) already excel at in-context learning, so why can’t we just feed tables into an off-the-shelf LLM?Because LLMs are trained on natural language rather than structured data, they struggle to process tables directly. First, their context limits are exhausted quickly by medium-sized tables containing just a few thousand rows and hundreds of columns. Second, LLMs suffer from tokenization inefficiency, awkwardly splitting numerical values and destroying mathematical precision. Finally, they suffer from structural blindness. When a 2D table is serialized as a 1D text string, LLMs lose track of which value belongs to which row and column as the table grows. “That’s why, today, it is far more effective to use an LLM to write the code that handles feature engineering and calls XGBoost than to ask the LLM to read the table itself,” Kong said.What is TabFM?To run inference with TabFM, you do not update any model weights. Instead, you take your historical examples (the training rows with their known labels) and your target rows (the new data you want to predict) and pass them to the model as a single, unified prompt. The model learns to interpret the relationships between columns and rows directly from this context at runtime.For example, consider an enterprise analyst trying to predict customer churn. Instead of building a bespoke data pipeline and training an XGBoost model, they can simply pass a sample of historical user session data alongside a new, active session into TabFM. In one forward pass, the model returns an instant churn probability. TabFM overcomes the limitations of LLMs by treating the data as a grid, preserving its structural integrity without forcing it into a single-dimensional text string.To effectively process diverse tabular structures while enabling scalable zero-shot prediction, TabFM synthesizes the strengths of earlier experimental architectures, TabPFN and TabICL. TabPFN, developed by Prior Labs, first proved that a transformer architecture could perform zero-shot classification on small tables, though it struggled to scale computationally to larger datasets. Later, TabICL, developed by France’s National Research Institute for Digital Science and Technology, addressed this bottleneck by introducing row compression, allowing in-context learning to efficiently process much larger tables. TabFM combines TabPFN’s deep feature contextualization with TabICL’s efficient compression into a novel hybrid design built on three key mechanisms:1. Alternating row and column attention: The raw table is first processed through a multilayer attention module that alternates across both columns (features) and rows (examples). By continuously attending across these two dimensions, the model natively captures complex feature interactions. This deep contextualization does the heavy lifting that would usually require tedious manual feature crafting by data scientists.2. Row compression: Following this contextualization, the cross-attended information for each row is compressed into a single, dense vector representation. TabICL pioneered this by using CLS tokens to compress a row’s rich information into one vector, “in contrast to TabPFN v2, v2.5, and v2.6, which attend over the full cell grid throughout the network,” Kong explained. This drastically shrinks the computational footprint.3. In-context learning (ICL): A causal Transformer then operates on this sequence of compressed embeddings. This Transformer model uses the attention mechanism of TabICL to attend over these dense row vectors, drastically reducing the computation cost and allowing the model to process large datasets efficiently.A major selling point of TabFM is its pretraining recipe. The model was trained entirely on hundreds of millions of synthetic datasets. These datasets were dynamically generated using structural causal models (SCMs) that incorporate a wide variety of random functions. By training exclusively on synthetic SCMs, TabFM learned the fundamental mathematical priors of how tabular features interact without ingesting real-world, confidential CSV files.TabFM in actionTo test the model’s capabilities, Google researchers benchmarked TabFM on TabArena, a comprehensive evaluation suite spanning 51 diverse tabular datasets across 38 classification and 13 regression tasks.On these public benchmarks, TabFM’s zero-shot predictions already match or beat heavily tuned supervised baselines. However, Google is careful to note that this does not automatically mean TabFM will universally dethrone bespoke, hyper-optimized production models on every enterprise workload.”Instead of replacing hyper-optimized production models, the true practical business value it unlocks for lean engineering teams is velocity,” Kong said. “It allows data analysts and backend engineers to instantly spin up high-quality baseline models without a dedicated data science team managing a complex lifecycle.”For advanced practitioners looking to squeeze out maximum accuracy, the research team also introduced a “TabFM-Ensemble” configuration. By running the model through 32 distinct variations and blending the results, TabFM pushes the performance even further. Getting started, trade-offs, and the cloud futureThe shift to in-context learning for tables introduces a new economic trade-off that engineering teams must consider. With traditional algorithms, training is slow and expensive, but inference is lightning-fast and cheap. TabFM flips this dynamic. While training time drops to zero, inference becomes significantly heavier. Because the model must process the entire historical dataset as context during every single prediction, it requires more compute and memory at runtime. In this new paradigm, “traditional machine learning training becomes the ‘prefill’ phase (KV caching) in the context window,” Kong said. While this prefill cost is steep, it is paid only once per table, and the cache is reused across subsequent queries. “The catch is prediction latency, which no amount of caching removes,” Kong added. Every new prediction requires a pass through a large transformer. “Any production API requiring single-digit-millisecond response times cannot tolerate TabFM’s forward-pass overhead.”For developers looking to evaluate the model today, the barrier to entry is low. Google designed TabFM as a drop-in replacement for traditional ML workflows, offering a scikit-learn compatible API (TabFMClassifier and TabFMRegressor). It natively handles mixed numerical and categorical columns, works directly with pandas DataFrames, and requires no manual ordinal encoders or numerical scalers. The library supports both JAX and PyTorch backends.However, enterprise teams need to be aware of current limitations and licensing restrictions. The model architecture has a hard limit of 10 output classes for classification tasks, and it is optimized for tables with up to 500 features. More importantly, while Google released the underlying codebase under the permissive Apache 2.0 license, the pre-trained model weights are published on Hugging Face under a strict tabfm-non-commercial-v1.0 license. Developers can evaluate the model internally, but it cannot be deployed in commercial products yet.Looking ahead, Google is addressing the commercial deployment friction through its cloud ecosystem. TabFM is being integrated directly into Google BigQuery, allowing analysts to run zero-shot predictions natively via an “AI.PREDICT” command. By putting foundation model inference right next to the data warehouse, TabFM could soon make complex tabular machine learning as accessible as a basic database query.In practice, TabFM shines in rapid prototyping, high data drift environments, and small to medium-sized datasets under 100,000 rows. Conversely, teams should stick to traditional models for strict, ultra-low latency APIs, or massive tables exceeding one million rows, which currently require aggressive row sampling that degrades the foundation model’s competitive advantage.

Despite President Donald Trump’s refusal to sign Congress’ bipartisan housing bill, it’s set to go into effect at midnight, and its temporary CBDC ban along with it.

Delta’s profit dropped as the air carrier absorbed the highest quarterly fuel expense in its history.

TheStreet aims to feature only the best products and services. If you buy something via one of our links, we may earn a commission.Why we love this dealOne of the best things about summer is that it finally gives us the chance to throw our winter boots to the side and break out the flip-flops. The only problem? After a while, your feet can take quite a beating in them. They’re comfortable and convenient, no doubt, but with little to no support and constant exposure to all Mother Nature has to offer, they’re not always the most optimal shoe choice. Thankfully, with more and more styles being released that are designed with additional support in mind, like the Coface Orthotic Flip-Flops, you can enjoy that loud thwacking sound you hear as your foot connects with the pavement every time you step and still keep your feet in tip-top shape. They’re specifically designed to offer more heel, arch, and sole support so that, even after a long day on your feet, your feet aren’t aching and in pain, and what’s even better is that they’re on sale for 25% off at Amazon. The $33 sandals are now only $25, and with so much of summer still to conquer, there’s no better time to add a pair to your cart. Coface Orthotic Flip-Flops, $25 (was $33) at Amazon

Courtesy of Amazon

Shop at AmazonWhy do shoppers love it?Orthopedic shoes in general utilize special features to relieve pressure, reduce pronation, and stabilize your joints. Some shoes have inserts and accessories added to them upon your purchase, whereas others are pre-designed with them in place so no work is required on your end, and that’s how these flip-flops are designed.The thong sandals have a bionic design to support high arches and a contoured footbed to hug your foot. With this in place, the strain on your feet is reduced, your stride is more stabilized, and easy injuries like plantar fasciitis or tendonitis are less likely to occur. The bottom portion of the shoe is made with ethylene-vinyl acetate (EVA) that’s a rubber-like copolymer similar to a foam. It’s lightweight and flexible, so you can move with ease while giving you shock-absorbing support that reduces the impact the ground makes on your foot with each stride. In fact, there’s even some memory-foam incorporated into the sole along with a heel cup to provide extra cushioning and stability while eliminating the risk of foot fatigue. The flip-flops have that standard thong design with an extra twist added. There is an additional strap that’s adjustable with a hook-and-loop fastener to fit your shoes more comfortably to your feet. It’s perfect for days where the heat makes your feet swell and you need a little bit more room than usual. The straps are made with faux leather, while the bottom exterior portion of the shoe has a non-slip textured rubber tread to provide traction and protection. There’s even an abrasion rubber sheet that provides stability in wet conditions so slick surfaces don’t pose a risk. Related: Skechers walking shoes that ‘feel like walking on clouds’ are just $40 at AmazonAvailable in 13 colors and in sizes 5 through 12, with half sizes in between, these sandals are great for casualwear with a pair of shorts and a T-shirt or dressed up for slightly more formal occasions with a pair of nice pants, a skirt, or a dress. Details to knowMaterial: Rubber, faux leather, and ethylene-vinyl acetate.Colors: 13. Sizes: 5 through 12 with half sizes.With over 11,500 five-star ratings, these flip-flops are certainly popular with shoppers. The contoured footbed provides really great arch support and feels super comfortable for the feet. The shoes are both functional and very attractive, and they are lightweight and perfect for the pool, beach, or running around town. “I’ve been having heel and plantar fasciitis pain so bad it was extremely difficult to get out of bed in the morning without limping,” one shopper said. “These are a game changer.”Shop more deals Cushionaire Double Buckle Slip-On Sandals, $45 (was $65) at AmazonKuaiLu Flip-Flops, $14 (was $23) at AmazonUbfen Hiking Sport Sandals, $40 (was $50) at AmazonOrthopedic shoes can be expensive, but you don’t always need to pay a fortune to get the technology and design that supports your feet best. Enjoy your favorite summer style with all the structure and support you need to live pain-free with the Coface Orthotic Flip-Flops. 

Online shopping is moving into a new phase, and consumers may not always notice who is powering it.A shopper today can ask an AI assistant to compare products, find deals, read reviews, build a cart, and move closer to checkout without visiting a traditional retailer’s website. In some cases, the assistant is still mostly helping the shopper search.In others, it is beginning to act more like an agent that can move the purchase forward with user approval, payment controls, and other guardrails.For Shopify, this distinction is becoming crucial as it heads to report its second-quarter 2026 results on August 5, before the market opens.Shopify, the commerce software company behind millions of online and in-store sellers, helps merchants run stores, manage products, process payments, and sell across different channels.Now, Bank of America is returning to the stock with a bullish view, arguing that the rise of AI shopping may not weaken Shopify’s role in online commerce. Bank of America returns to Shopify Bank of America reinstated coverage of Shopify with a Buy rating and a $150 price target in a July 7 note reviewed by TheStreet.The firm said that the company could be a “core beneficiary” of the move toward AI-driven, agentic commerce.BofA analyst Tal Liani said investor concern has centered on whether AI shopping tools could bypass merchant websites and shift discovery and transactions away from Shopify’s platform.More AI:The new Chinese AI model rattling U.S. tech investorsAnthropic restores access to Mythos 5 for select organizationsSoftBank CEO offers stinging critique of Musk’s AI betThis worry has been weighing on the stock, which is down around 24% year to date.BofA’s $150 price target implies nearly 25% upside from the stock’s price of $120 at the time.In the reinstatement note, Liani argues that investors may be looking at the wrong part of the shopping journey.If AI assistants become the new front door for product discovery, the value may shift toward the systems that enable transactions.This includes product catalogs, real-time inventory, pricing, checkout, payments, and fulfillment. Those are areas where Shopify is already embedded across its merchant base.BofA’s insight gives investors an early framework for what to watch: not just whether Shopify beats quarterly expectations, but whether its AI, payments, international, and enterprise growth story is gaining strength.

Shopify’s stock is down 25% year to date.Iryna Tolmachova / Getty Images

AI shopping has reached checkoutAI shopping is no longer a theoretical concept.Amazon recently introduced Alexa for Shopping, which combines Rufus and Alexa+ to help customers compare products, track price history, build carts, and reorder essentials.For eligible products, consumers can use Amazon’s Buy for Me agentic AI feature to shop across the web.Walmart and Google are also pushing AI shopping closer to the checkout process. Google’s Gemini shopping expansion includes partnerships with Walmart, Shopify, Wayfair, and other retailers, allowing shoppers to find products and, in some cases, buy without leaving the Gemini chat.Microsoft is adding a similar layer to Copilot. The company’s Copilot Checkout feature can show a buy option inside an AI conversation, then open an in-chat checkout flow where shoppers enter shipping and payment details and confirm a purchase. Microsoft’s initial retail partners include Urban Outfitters, Anthropologie, Ashley Furniture, and some Etsy sellers, while PayPal, Stripe, and Shopify are working with Microsoft on payments.For merchants, this change is big.If shoppers increasingly buy through AI surfaces, merchants need their product data, inventory, pricing, and checkout systems to work inside those interfaces.But Shopify isn’t just waiting to see how AI changes online retail.It is helping build some of the infrastructure that could support it.The company already holds a big 14%+ share of total US e-commerce, second only to Amazon. To leverage this scale as consumer behavior shifts, Shopify has co-developed the Universal Commerce Platform (UCP), an open-standard technical framework built alongside major technology and payments companies, including Amazon, Meta, Microsoft, Stripe, and Salesforce. The UCP serves as a critical strategic lever by standardizing exactly how autonomous AI agents discover inventory, negotiate terms, and complete transactions across the web. By aligning with such companies, Shopify is effectively positioning its backend ecosystem to remain important as AI becomes an increasingly important part of digital commerce.Shopify earnings will test AI-commerce caseShopify’s first-quarter results already gave Wall Street a strong base for the AI-commerce debate.The company said revenue rose 34% in the quarter ended March 31, 2026, while free cash flow margin was 15%. Gross merchandise volume, the total dollar value of orders facilitated through Shopify’s platform, cleared $100 billion in the quarter, reaching $100.7 billion.Shopify also told investors it expected second-quarter revenue to grow at a high-twenties percentage rate year over year.It expects gross profit dollars to grow at a mid-twenties rate, operating expenses to be 35% to 36% of revenue, and free cash flow margin to be in the mid-teens.BofA’s longer-term model is more bullish than a single quarter. The firm expects Shopify revenue to grow 24% to 28% annually from fiscal 2026 through fiscal 2028, supported by three drivers: agentic commerce, international expansion, and larger enterprise merchantsThe AI data in the note is especially important because it shows the shift is already showing up in Shopify’s merchant activity. BofA said AI-driven traffic to Shopify merchants rose eightfold year over year in the first quarter, while orders from AI-powered searches increased about 13 times. New-buyer orders from AI surfaces were occurring at nearly twice the rate of traditional channels.BofA also highlighted Shopify’s Catalog product, which feeds inventory and pricing into AI agents, and Sidekick, Shopify’s AI assistant for merchants. The firm said traffic from Catalog-powered AI searches converted to purchases at twice the rate of general AI search traffic, while Sidekick weekly active users were up fourfold year over year.The point is not that every shopper is suddenly handing purchases to a bot. Most AI shopping systems still require confirmation, limits, or other controls before money changes hands. The bigger shift is that checkout is moving closer to the AI conversation itself.That could make Shopify less visible to consumers, but more important to merchants that need to be available wherever shoppers start.Shopify has more than one growth driverBofA’s call on Shopify is not limited to AI.International growth is another key part of the firm’s bullish case. BofA said international gross merchandise volume for Shopify rose 45% year over year in the first quarter, faster than overall GMV growth. The firm also said Shop Pay’s gross merchandise volume outside the U.S. grew more than 70% year over year.This is important because Shopify, which is already a major U.S. commerce platform, still has room to expand globally as more merchants adopt localized payments and cross-border tools.Shopify’s move upmarket is also important. The company built its reputation with small and midsize merchants, but larger sellers are now becoming a bigger part of the story.BofA said merchants with more than $25 million in GMV were Shopify’s fastest-growing cohort.The number of merchants doing more than $100 million in GMV doubled over the past two years. Shopify Plus’s monthly recurring revenue grew 20% year over year and represented 35% of total monthly recurring revenue, according to the note.That enterprise shift could make Shopify’s revenue more durable if larger merchants increasingly use its payments, checkout, point-of-sale, and commerce tools over time.Shopify faces AI and competition risksThe bullish case, however, is not without risks.BofA said downside risks to its $150 Shopify price target include greater-than-expected disintermediation from AI-native commerce platforms.A slower adoption of key merchant solutions, execution risk in international expansion, and increased competition as Shopify moves upmarket.Those risks matter because AI could still change the balance of power in online retail. If major AI platforms, marketplaces, or payment companies control more of the shopping journey, Shopify will need to prove that its infrastructure remains essential.For investors, August 5 will be the next checkpoint. Shopify’s Q2 earnings will show whether the company can keep posting strong growth while defending its place in the online shopping stack.For shoppers and merchants, the bigger question is direct: when AI changes how people buy, will Shopify become less visible, or more powerful behind the scenes?Related: Costco quietly makes a key credit card change

The South Korean memory company’s American depositary receipts begin trading on Friday.

Trump Says Iran Peace Talks Will Continue Despite End To Ceasefire

Essential guidance for daily fantasy sports bettors, with emphasis on 2026 NFL. Difference between cash games and GPP, know it live it and score big this season!

Micron Technology Inc. (MU) said it will raise its planned U.S. investment to more than $250 billion through 2035, according to a Seeking Alpha report on the company’s Thursday, July 9, announcement.That figure is $50 billion more than the roughly $200 billion the company committed to just over a year ago, based on a Micron SEC filing.The increase arrived alongside a construction milestone in Clay, New York, where Micron poured the first concrete for its new megafab three months ahead of schedule.The timing matters more than the number. Micron isn’t expanding because it wants more capacity someday. It’s expanding because it can’t build fast enough to keep up with demand it already has, and its own CEO won’t say when that pressure eases.Micron flags memory shortage with no end date attachedIn an interview with FOX Business’s Liz Claman, Micron chairman and CEO Sanjay Mehrotra said “memory is in deep shortage right now,” and that the expanded investment is meant to pull in the timelines on new supply.It’s a direct admission that the $250 billion figure is defensive as much as ambitious.Claman pressed him on when the shortage would end, noting Micron had previously said tightness would last beyond 2027. Mehrotra wouldn’t commit to a date.Related: Veteran analyst drops massive Micron valuation prediction“We are not putting a date or month on it, because the demand just continues to go up as well,” he said in the same interview.Part of that is simple physics. Mehrotra told Claman that “from shovel in the ground to getting first silicon out is good three to four years’ time frame,” even when construction moves fast.That timeline is why Micron is pouring concrete now for supply it won’t ship until later in the decade.That uncertainty is backed by contracts, not just talk. Mehrotra said Micron has signed 16 customers to strategic supply agreements running as far out as 2030, a sign buyers expect the crunch to outlast this investment cycle.

Micron raised its U.S. investment plan to $250 billion as CEO Sanjay Mehrotra declined to say when the memory shortage will end.Bloomberg / Getty Images

Wall Street reads Micron rally as more than optimismMicron (MU) shares climbed almost 5% on July 9, according to CNBC. Other chip equipment and design names, including Applied Materials, KLA, Lam Research, and Arm Holdings, rallied the same day.That breadth matters. Investors weren’t just repricing Micron. They were pricing in a longer AI infrastructure buildout across the memory supply chain.The move follows a separate bullish signal. Citi analysts placed Micron on an upside catalyst watch this week, citing expectations that DRAM prices could nearly triple in 2027. Combined with the July 9 investment news, that forecast suggests Wall Street sees this shortage as a multi-year pricing story, not a short squeeze.The rally still comes with a caveat memory investors know well. Micron cut about 15% of its global workforce in 2023 when memory prices collapsed during the last downturn, a history Claman raised directly in the interview.A $250 billion, decade-long bet only pays off if this cycle doesn’t repeat that one.Apple is already paying for the memory shortageMicron doesn’t operate in isolation, and the same tightness fueling its investment plans is squeezing its customers.Apple raised prices on iPads, Macs, and other hardware by roughly $100 to $200 per device in late June, citing what it called an unprecedented jump in memory and storage costs. Apple shares fell as much as 6% the day the increases took effect.More Micron:Morgan Stanley resets Micron stock price target on strong AI demandMicron just dethroned Nvidia in one key wayBank of America strongly resets Micron stock price targetReuters reported that memory makers including Micron have prioritized orders from AI chipmakers like Nvidia in recent months, leaving less supply for consumer electronics makers.That dynamic helps explain why Micron can justify quadrupling domestic capacity. Its highest-value customers now are the AI buildout itself, not the phone and laptop makers that used to set the terms.The bigger story is where the chips get madeBeyond the dollar figure, Micron’s plan is a bet on geography. The company wants 40% of its DRAM output made domestically, and the Clay campus is expected to generate 50,000 New York jobs, including 9,000 direct roles at Micron.Micron is also putting up to $3 billion into supply chain partners, including $500 million in financing for GlobalWafers’ Texas wafer facility, tying a second country’s raw silicon supply into its U.S. footprint through a new 10-year agreement.Washington has framed the deal in strategic terms.U.S. Commerce Secretary Howard Lutnick called the expanded investment a matter of national security in comments cited by the Seeking Alpha report, tying memory manufacturing to broader technology leadership.That framing, reshoring production while locking down raw material access, reflects a wider shift in how chipmakers are hedging against both AI demand spikes and geopolitical risk.Micron isn’t the only company making that bet, but it’s making one of the largest and most public versions of it.The open question isn’t whether Micron builds these fabs. Construction is already ahead of schedule. It’s whether the shortage driving this spending holds long enough to justify it. While Mehrotra won’t put a date on that, investors will need to watch Micron’s upcoming earnings report closely for any shifts in the memory demand cycle.Related: Tokyo puts billions behind Micron’s chip plan

The initiative builds on a broader shift by companies exploring stablecoins to move money between international operations more efficiently.