- Tech Stock Insider
- Posts
- The AI Hardware Shakeout: Four Tech Names To Watch After The Selloff
The AI Hardware Shakeout: Four Tech Names To Watch After The Selloff
Rate-hike fears and a wave of downgrades hit AI hardware this week. One software name went the other way.
AI infrastructure stocks just got hit hard. A hot inflation print revived Fed-hike bets, analysts started downgrading hardware names, and storage, server and fiber stocks are now 15% to 45% off their highs.
Four names, very different setups, each with a catalyst close enough to matter. You get the thesis, the damage, and the risk on each.

Elite Picks (Sponsored)
This report focuses on a narrow group of stocks identified through a detailed screening process.
Analysts apply a combination of metrics to narrow down potential opportunities.
Past selections have shown strong momentum, but no outcomes are guaranteed.
The newest edition is now open for access.
Get the report now.
*This free resource is being sent by Zacks. We identify investment resources you may choose to use in making your own decisions. Use of this resource is subject to the Zacks Terms of Service.
*Past performance is no guarantee of future results. Investing involves risk. This material does not constitute investment, legal, accounting, or tax advice. Zacks Investment Research is not a licensed dealer, broker, or investment adviser.

Foundation Models
Microsoft Completes Pretraining of a 2.8 Trillion-Parameter AI Model

Microsoft (NASDAQ: MSFT) has completed pretraining of MAI-Base-2, an in-house AI model containing 2.8 trillion total parameters. The company says the model surpassed every open-weight system included in its internal testing.
MAI-Base-2 uses a mixture-of-experts architecture that activates only 80 billion parameters for each task.
Selective activation allows the model to draw from a much larger system without using all 2.8 trillion parameters every time it produces a response.
Built on 50 Trillion Training Tokens
Microsoft trained the model using 16,000 NVIDIA GB200 GPUs over 38 days. Its training dataset contained roughly 50 trillion tokens, with the amount of unique material growing 36 percent from the earlier MAI-Base-1 model.
The dataset also included original material rather than knowledge distilled from competing AI systems. Images accounted for 18 percent of the training mixture, giving MAI-Base-2 native visual capabilities alongside text processing.
Efficiency Becomes the Main Claim
Microsoft says its training approach required between 4.2 and 12.3 times less computing power than DeepSeek-V4-Pro-Base across four evaluations.
Its updated training process was also between 6.3 and 12.7 times more compute-efficient than MAI-Base-1.
MAI-Base-2 is now entering post-training, where Microsoft will refine its behavior and specialized capabilities. Microsoft has not announced pricing or commercial availability, but work on MAI-Base-3 is already underway.

Software
Google Launches Universal Cart Across Search and Gemini

Alphabet (NASDAQ: GOOGL) has launched Universal Cart across Google Search and the Gemini app in the United States.
The new system stores products from different retailers inside one persistent cart instead of separating them across websites, browser tabs, and Google services.
Products added through AI Mode in Search remain available when shoppers later open Gemini. Universal Cart uses the Universal Commerce Protocol to connect product discovery, cart management, and checkout across participating retailers.
The Shopping Graph Works in the Background
Universal Cart connects to Google’s Shopping Graph, which contains more than 60 billion product listings that are continually refreshed. The system monitors saved products and sends notifications when prices drop or unavailable items return to stock.
Google also uses AI reasoning to identify promotional codes and compare benefits attached to payment cards stored in Google Wallet. Relevant savings appear directly in the cart, while products from different merchants stay organized together.
Retailers Stay Inside the Transaction
Users can complete supported purchases through Google Pay without leaving Universal Cart. Products can also be transferred directly to a retailer’s website when the merchant handles checkout through its own system.
Nike, Target, Walmart, Sephora, Wayfair, Ulta Beauty, and several Shopify merchants support these simplified purchasing options.
Retailers remain responsible for sales, confirmations, and customer support, while Google provides the software that connects the shopping process.

Big Money Moved (Sponsored)
The size was never the problem. The reach was.
Then a drilling crew hit the DOE's 2035 targets twelve years early, and costs fell 50% in 18 months.
Google signed a 15-year contract.
Bill Gates wrote a $100 million check.
The Pentagon made it a priority.
One company has quietly built this for sixty years, and owns the entire chain.
See the Grand Canyon energy stock behind this federal advantage

Smartphones
Apple Takes the iPhone Into Foldable Hardware

Apple (NASDAQ: AAPL) is opening preorders for iPhone Duo, its new foldable smartphone built around the company’s largest iPhone display.
The inner screen measures 7.58 inches and provides 50 percent more display space than the iPhone 18 Pro Max.
When folded, iPhone Duo uses a compact 5.4-inch outer display that retains 90 percent of the screen area found on iPhone 18 Pro. A precision-engineered hinge connects both sides and keeps the phone thin when fully opened.
New Hardware Runs Beneath the Hinge
Apple’s A20 Pro chip powers the device, while a dual-battery architecture distributes power across its folding body. The phone also carries an IP68 rating for water and dust resistance despite the additional moving components.
Its camera system includes 48-megapixel main and ultra-wide cameras, an optical-quality 2x telephoto option, and an inner FaceTime camera.
The device can remain partially folded and support itself during video calls, photography, or media playback.
iOS Learns a New Shape
Apple redesigned iOS 27 to react as iPhone Duo opens, closes, rotates, or switches between its two displays.
Split View lets two applications run side by side, while StandBy can run on either screen without requiring the phone to be charging.
Siri AI can stay visible beside an app and use on-screen information to answer questions or complete actions.
Apple has turned the foldable display into a new software environment instead of treating it as a larger version of the existing iPhone screen.

Would you rather a tech giant spend its cash on one massive acquisition, or spread the same total across dozens of smaller ones? |

Recent Tech Movers
Everpure (NYSE: P)
AI Data Hunger Pulls Storage Higher
Everpure (NYSE: P), the company that rebranded from Pure Storage in February 2026 and switched its NYSE ticker from PSTG to P this spring, is one of the overlooked beneficiaries of the AI buildout you should know.
Every model checkpoint, every enterprise data pipeline, every training dataset has to live somewhere, and increasingly that somewhere is an all-flash array feeding GPU clusters.
Risk lies with a valuation reset. At that multiple, any slip in bookings cadence or margin pressure from product mix hits the stock hard, as this week showed.
The next real test is the fiscal third-quarter report, expected around November 18, and the pace of large deal bookings.
Flex Ltd. (NASDAQ: FLEX)
The Servers These Models Actually Live On
Flex (NASDAQ: FLEX) is the kind of name you ignore because it sounds boring. Contract manufacturing. Assembling things other companies design.
But over the past 18 months, the company has repositioned itself as a core AI infrastructure partner, building liquid-cooled server racks, power modules, and data center hardware for the hyperscalers spending hundreds of billions a year on compute.
One catch: Flex plans to spin that Cloud and Power Infrastructure business off as a separate company, Axiom Solutions, with Flex keeping only a 6% to 12% stake.
Cycle timing is the risk. If AI CapEx normalizes next year or a single large customer pulls back orders, Flex's growth decelerates fast. Earnings land October 28, and any update on the Axiom spin-off terms will matter as much as the quarter.
Ciena Corporation (NYSE: CIEN)
Fiber Demand Catches the Bid
Ciena (NYSE: CIEN) became an AI play when hyperscaler orders for optical networking gear went vertical. Every multi-site AI cluster needs high-speed coherent links to move training data between buildings, and Ciena is the only pure-play optical systems provider operating at scale today.
The stock has been a roller coaster. It jumped 13.8% on Tuesday, October 6, after upbeat commentary from Marvell and Nokia, then gave all of that back and more.
If you want secular growth exposure tied to the data center buildout, the pullback lets you buy it well off the top, at about 50 times forward earnings and a market cap near $51 billion.
With $8.5 billion in undelivered orders at quarter end and Ciena Ventures, a $200M corporate venture arm focused on AI-driven networking and data center technologies, this is one of your cleaner ways to own the buildout.

Sell Rating Alert (Sponsored)
It could be in your 401(k) anchoring your portfolio.
But our independent Weiss Ratings, which have correctly called nearly every major financial event of the 21st century, just slapped this popular stock with a "SELL".
And it's not the only one.
We found nine other popular but toxic stocks.
Click here to discover the 10 toxic stocks and protect your wealth now

The AI Supply-Chain Story Hiding in Plain Sight
Manhattan Associates (NASDAQ: MANH)
Why it made the radar
Manhattan Associates (NASDAQ: MANH) is a supply chain software vendor you probably haven't heard of, which is part of what makes it interesting right now. The company sells warehouse management, order management, and transportation planning software to the biggest retailers and 3PLs in the world.
More importantly, it's one of the earliest enterprise SaaS vendors to actually embed generative AI into its core workflows, not as a bolt-on but inside the Active Omni platform itself. Store associates and warehouse operators can now query inventory, routing, and fulfillment in natural language.
The setup and the risk
This is the name that went the other way. While AI hardware sold off, Manhattan closed Thursday at $226.96, a 52-week high, up about 36% this year, with a market cap near $13 billion.
The next test is Q3 earnings on Tuesday, October 27, after the close. If cloud revenue growth holds up and the AI features show up in new deals, the breakout has room. The risk is competitive.
Oracle and SAP are both pushing hard into the same AI-for-supply-chain pitch with deeper pockets. If Manhattan can't defend its install base, the multiple (about 35 times forward earnings) compresses before the thesis has time to play out.

Everything Else
📊 Great stocks tend to look boring right when they matter most, and our analysts flagged 7 still building scale outside the spotlight heading into 2026.
🤖 ServiceNow launched its AI Passport platform, giving businesses greater control over AI agents' access to sensitive data and helping shares outperform the broader tech selloff.
📡 Ericsson surged on stronger quarterly results, gaining nearly 7% as improved network sales and cost reductions lifted profitability.
💾 Semtech raised its revenue outlook to approximately $417.5 million for its fiscal third quarter, while also increasing its adjusted earnings guidance.
💻 Intuit jumped more than 6% during Thursday's session as enterprise software stocks rebounded despite continued weakness among semiconductor companies.
🖥️ Aixtron plunged 13% in Thursday's trading, leading European technology losses as concerns about AI infrastructure spending weighed on chip-equipment suppliers.

That's our coverage for today; thanks for reading! Reply to this email with feedback or any tech stocks you want me to check out.
Best Regards,
—Noah Zelvis
Tech Stock Insider


