The clearest sign that AI is changing from a promise into infrastructure is what companies are now selling. They are not selling smarter chatbots. They are selling locks, data pipelines, chips and sales channels. Between 20 October and 6 November 2026, Salesforce agreed to buy Braze for about $5.66 billion in enterprise value, or $47.50 in cash per Braze share.1 Fortinet launched a service to inspect and guard AI traffic.2 A robotics-data firm described the cost of feeding physical AI.3 Everything below comes from company announcements and executives' own words, so it shows what the industry says it is doing, not independent proof that it works.
A note on how far to trust this
Nearly all of the material here comes from press releases carried by a wire aggregator, NewsEOD. Via News's own measurement found that 54% of 6,613 claims checked from that source held up.4 The Crunchbase piece on robotics funding scores lower, 49%, but only 41 of its claims were checked, so that figure is rough.5 Treat the numbers below as what the companies reported. Where we can add arithmetic from those numbers, we say so.
Securing AI is becoming a product category
Fortinet announced the FortiGuard Inline AI Security Service on 26 October. It is a cloud-delivered service that adds AI-specific inspection and guardrails to FortiGate firewalls.2 The pitch is that companies do not need new equipment. Michael Xie said: "By extending advanced AI inspection and guardrails through FortiGate, we are making AI security available through the infrastructure customers already use."2 The company also says customers can choose how to deploy and use the security based on cost, performance, privacy and sovereignty requirements.2
Cisco is making a similar bet through its sales partners. It announced general availability of Cisco IQ for partners and expanded Cisco Cloud Control offerings meant to help partners build, secure and manage AI-ready infrastructure.6 Tim Coogan said: "Our partners are where Cisco meets the customer, and the AI era is asking more of all of us."6
T-Mobile supplies the closest thing to a result. It published a report called Advancing Cybersecurity at T-Mobile on how it rebuilt its security program.7 Jeff Simon said: "Today, phishing-related credential compromises among our people are down more than 99%."7 That is a strong claim, and it is T-Mobile's own. The report is described as covering what held up when tested, including bug bounty results, but the figures from those tests are not in our material.7
Physical AI: the shortage is data, and data is costly
Crunchbase reported that global venture funding to robotics startups reached $9.9 billion in Q3 2026, more than double the $4.3 billion of Q3 2025.5 To picture that scale, it is nearly twice the roughly $5.66 billion Salesforce agreed to pay for Braze.1,5 The bulk went to companies in China and the US.5
Ulrik Stig Hansen of Encord explained why the money is going into data. He said: "Physical AI is very different because there is no internet of physical AI data, and so the data first has to be collected, and you have to convert it from sensor tokens to action tokens."5 In his words, "storing all that data, curating all that data, enriching all that data, and validating that all the data works for these systems is a very expensive operation."5 He also described the industry as still in the pre-training phase of physical AI, though it will soon enter the post-training phase.5
Hardware makers say the other bottleneck is the chip inside the machine. At the 9th China Robot Summit, Harvest Praxis launched the "Hyper AI Brain". It is an on-device computing platform built on MediaTek's 3nm chip, and the company says it is the first to fuse a robot's perception and decision-making with its motion control on one chip.8 Philip Cai Zhenhui said: "The industrialization of embodied intelligence relies on three critical pillars: large models, real-world data, and on-device computing."8 Harvest Praxis calls on-device computing the most fundamental yet weakest link.8 The company is an interested party, since it sells that chip platform.
Consolidation in customer engagement
Salesforce's agreement to buy Braze, a customer engagement platform listed as NASDAQ: BRZE, is the largest deal in our material.1 Braze coordinates personalised messages across mobile apps, websites, email, SMS and WhatsApp.1 Braze's Bill Magnuson said the company "shipped at the fastest pace in our history" this past year and that BrazeAI helps marketers get ideas live faster.1 The reading of this deal as consolidation toward AI-driven customer engagement is our interpretation. The press release does not say it.
The small end of the market: TEN Holdings
TEN Holdings is a Nasdaq-listed company (XHLD) whose subsidiary runs a Digital Experience Platform. It shows how the shift looks away from the big names.9 Third-quarter revenue rose 17.9% year over year to $640,000.9 The same results include a $1.8 million write-off of capitalised service contracts and about $0.5 million of legal expenses.9 The write-off alone is nearly three times the quarter's revenue, which is our arithmetic from the reported figures.
Chief executive Virgilio Torres said the company cut its annual operating cost base by about $2.5 million over 12 months, through "strategic cost reductions, process improvements, and the integration of AI-driven tools."9 Short-term related-party borrowings fell from $4.6 million to about $0.1 million at 30 September 2026. Total liabilities fell about 74%, from $7.7 million at the end of 2025 to $2.0 million.9 The dossier does not include the company's net loss or cash position, so we cannot say whether the turnaround is working. A smaller balance sheet and a cheaper cost base are not the same thing as a profitable business.
Devices and the edge
On the consumer side, Apple unveiled a new iPad mini with an 8.5-inch OLED display and an A20 Pro chip.10 Pre-orders opened on 28 October and it became available on 6 November.11 Our material has no specifics on NVIDIA's RTX Spark laptops, so we do not describe them.
Impinj, which makes RFID chips used to tag and track goods, is a useful reality check for hardware. It reported record third-quarter revenue of $112.1 million. Chief executive Chris Diorio credited retail demand and new products and said the trends are continuing.12 Yet its own fourth-quarter revenue guidance is $104.0 to $107.0 million, below the record quarter.13 Hardware demand is strong, but not rising in a straight line.
What to watch
- Evidence behind security claims. T-Mobile's 99% figure is a company statement. Look for independent testing of AI-security products like Fortinet's.2,7
- Whether post-training arrives. Encord's executive expects physical AI to move to its next phase soon. The cost of data is the thing to track.5
- Braze deal progress. The deal is announced, not closed.1
- TEN Holdings' next report. The test is whether revenue grows while the write-offs and legal costs stop.9
- Impinj's fourth quarter. It will show whether record demand holds against lower guidance.12,13
The theme running through every source is that executives are upbeat, and the hard part is no longer making AI models. It is securing them, feeding them data and paying for the integration. That is our synthesis of what these executives said. No single source states it in those words.
