AI to ROI News & Analysis: June 19, 2026

Anthropic’s top models pulled from market; SpaceX closes the Cursor deal and has a great first week as a public company; OpenAI burns $38.5B; chatbot market changes, and more

The Biggest AI News This Week

🚨US Government Pulls Anthropic’s Most Powerful Models in Historic Export Control Order, Creating Chaos for Anthropic, Enterprises, and Governments Globally

🚀 SpaceX Closes $60B Cursor Acquisition in the Largest M&A Deal Ever for a VC-Backed Company

🤖 Microsoft Considers DeepSeek as a Lower-Cost Option for Copilot Cowork

💸 OpenAI Lost $38.5 Billion in 2025 as Costs Grew 8x Over the Prior Year

🔒 Databricks Acquires Panther Labs, Pushing Into AI-Powered Cybersecurity

🤝 OpenAI Launches Partner Network with $150M Investment, Targets Creating an Ecosystem of 300,000 Certified Consultants

📊 ChatGPT’s Market Share Slips Below 50%, with Big Gains for Gemini and Others

🛍️ Salesforce Acquires AI Customer Service Vendor Fin for $3.6 Billion

📈 SpaceX Has a Strong First Week as a Public Company

📜 This Week’s Great Report: The State of AI 2026 by Sensor Tower

📊 Very Useful Data: AI versus People, IPO Size, & the Effect of AI Spending

🃏 Definitely NOT AI: Cast iron tourist binoculars. Your Dad’s Nest camera. The best free restaurant bread. A snail’s promise. Nano Banana & the Mona Lisa.

1️⃣ US Government Pulls Anthropic’s Most Powerful Models in Historic Export Controls Order, Creating Chaos for Anthropic, Enterprises, and Governments Globally

The world's most powerful AI models went offline on June 13 after the US Department of Commerce ordered Anthropic to suspend all access to Fable 5 and Mythos 5 for any foreign national, both inside and outside the United States. This includes Anthropic’s own employees who are foreign nationals. Anthropic complied by disabling both models for all customers.

Issued via an “is informed” letter from Commerce Secretary Howard Lutnick, the directive cited national security authorities and threatened criminal and civil penalties for non-compliance. It is the first time US export controls have been applied to a domestically developed AI model. Negotiations between Anthropic and the Trump administration are ongoing, led by the Commerce Department and National Cyber Director Sean Cairncross.

How Anthropic Lost the US Government’s Trust

The dispute traces to a sequence of missteps over access to Mythos:

• Anthropic gave the government a list of 111 organizations cleared for early access to Mythos, then quietly expanded it by ~50 more.

• When the US government finally received the full list, one recipient, the South Korean telecom company SK Telecom, is suspected of having ties to China.

• Once alerted, Anthropic revoked SK Telecom’s access to Mythos, but the episode badly damaged officials’ confidence.

• The Commerce Department, CIA, and NSA all pushed to move ahead with export controls.

• Amazon CEO Andy Jassy then called Treasury Secretary Scott Bessent directly to report that Amazon researchers had found a way to bypass Fable 5’s guardrails, providing the administration with the final justification it needed.

• Administration officials gave Anthropic ninety minutes to take the models offline.

• When Anthropic did not immediately comply, the Commerce Department issued its export controls directive.

The Two Sides of the Dispute

The government’s position

• Fable’s guardrails can be bypassed, and Anthropic was “irresponsible” in resisting government security requests.

• Former AI Czar David Sacks said the administration acted “reluctantly” and that Anthropic had “prioritized its consumer model over safety.”

• One official said Anthropic “dug their own grave.”

• Trump personally signed off on the export controls despite reservations that they would hinder innovation, a senior White House official said.

Anthropic’s position

• The vulnerabilities are minor findings also present in other publicly available models, including OpenAI’s GPT-5.5.

• Cybersecurity expert Katie Moussouris of Luta Security, who reviewed Amazon’s findings at Anthropic’s request, said the behavior flagged was “the model working as intended” for cyberdefense, not a jailbreak.

• More than 150 security professionals, including former NSA and Facebook security leaders, signed an open letter calling for the controls to be lifted.

Geopolitical Fallout

The fallout isn’t limited to the United States:

• The European Commission opened an assessment of EU AI policy.

• Canadian Prime Minister Mark Carney cited the episode as proof of the risk of “overreliance on certain models.”

• G7 country access is explicitly off the table for now, with a Trump official calling any exemption for allied nations “completely illogical.”

• Dean Ball, a former Trump AI adviser who will join OpenAI in a policy role in a few weeks, wrote that the controls amount to a de facto AI licensing regime, which directly contradicts the President’s June 2nd executive order declaring that the US would not create a mandatory permitting requirement for AI.

• The export controls also prevent the NSA from using Mythos for cyber defense, creating a potential national security gap that the administration has not addressed.

The Takeaway

• The US now has a de facto global AI licensing regime, with no consistent rules. The Trump administration is applying export controls to one company’s AI products without providing a clear justification. Enterprises and governments building on US frontier AI must now factor in the risk that their models can be switched off at any time, for any reason, if the government believes a product creates a national security concern.

• Amazon’s role is an under-covered story. Amazon is Anthropic’s largest investor and primary cloud provider. Its internal security research triggered a US government response, leading to the shutdowns of Mythos and Fable. Did Amazon reach out to Anthropic first with its concerns and not get a satisfactory response? Or did CEO Andy Jassey simply reach out to the Secretary of the Treasury to score political points? We don’t know, but it does provide some intrigue.

• A resolution is possible but doesn’t seem imminent. Both sides say they want a quick fix. The most likely path is a limited technical fix to Fable’s guardrails, government verification, and a phased restoration, with US users returning first, followed by allied nations over a period of weeks. A full impasse scenario – models offline for months – would materially damage Anthropic’s business.

• This precedent will affect the launch of every frontier model going forward. Anthropic’s competitors should be concerned: Every frontier model can be used to launch highly effective offensive cybersecurity attacks. Therefore, model makers should assume that the US government may halt or limit the distribution of their products due to cyber concerns. Model makers should also assume that the voluntary pre-release model review process established by the President’s June 2026 EO is now effectively mandatory.

• Anthropic gets an “F” in geopolitics. The other big tech companies employ geopolitical savants who usually resolve disputes with government bureaucrats to their benefit with little or no drama. At Microsoft, it’s Brad Smith. At Apple, it’s Tim Cook. At Meta, it’s Dina Powell. At Anthropic, it’s DEFINITELY NOT Dario Amodei. Here are a couple of examples:

  • Anthropic is currently designated as a supply chain risk by the US Department of War. The other six major AI labs agreed to product-use terms very similar to what Anthropic wanted and now have large contracts in place. Anthropic is still living in the Litigation Hotel.

  • Anthropic selected the companies that have access to its Mythos model. Why didn’t it get US government approval first – avoiding objections and future punishments?

The most important short-term hire for Anthropic needs to be its version of Brad Smith. Things like this shouldn’t be happening to the world’s fastest-growing and most technologically advanced AI company.

2️⃣ SpaceX Closes $60B Cursor Acquisition in the Largest M&A Deal Ever for a VC-Backed Company

Four days after its record-breaking $75 billion IPO, SpaceX announced it was exercising its option to acquire Cursor parent Anysphere for $60 billion in an all-stock transaction. The deal is the largest-ever acquisition of a VC-backed company. It is expected to close in Q3 2026. Cursor is generating close to $4B in annualized revenue and serves 60% of the Fortune 500. Its lack of access to compute to train frontier models had “bottlenecked” its growth. SpaceX provides it with almost unlimited compute access. SpaceX stock climbed an additional 5% on the announcement, pushing first-week gains to more than 50% from the IPO price of $135.

What SpaceX Gets

The acquisition directly addresses SpaceX’s most significant competitive weakness, a hit AI product:

• Its Grok AI model has lagged Anthropic’s Claude Code and OpenAI’s Codex in enterprise coding benchmarks.

• Cursor provides proven enterprise distribution, a $4 billion annualized revenue run rate (per Forbes), and developer loyalty that Grok has not earned.

• SpaceX said it will soon release a joint AI model on Cursor alongside Grok Build, xAI’s coding agent, which the two companies have been co-training since April.

• Cursor CEO Michael Truell said the deal would advance Cursor’s “frontier AI capabilities with the goal of building the world’s most useful AI models.”

“One of the things that makes SpaceX so valuable is how valuable it is,” noting that a $60 billion all-stock deal costs materially less in dilution at a $2.5 trillion market cap.

Matt Britzman of Hargreaves Lansdown noted that “Cursor does not have the scale of OpenAI or Anthropic, but it has built some very impressive coding models relative to cost.”

The Takeaway

• SpaceX used its IPO currency brilliantly. All-stock deals at peak valuation multiples are how dominant companies buy fast-growing targets cheaply with minimum shareholder dilution. The Cursor deal follows the same playbook that Google used to acquire YouTube and Facebook used to acquire Instagram. Whether it works depends on SpaceX’s ability to successfully integrate and keep the Cursor development team.

• Cursor solves SpaceX’s enterprise credibility problem. xAI never effectively penetrated the Fortune 500. Cursor already has. The question is whether SpaceX can maintain Cursor’s rocket-ship growth trajectory while improving Grok sufficiently to replace OpenAI’s and Claude’s roles as Cursor’s underlying models.

• This is the first major test of the public company SpaceX. Post-close execution will be the real test of whether the transaction is worth $60B.

• The deal reshapes the AI coding market. Anthropic (Claude Code), OpenAI (Codex), and now SpaceX (Cursor + Grok Build) all have major enterprise coding offerings. The competitive pressure on pricing and performance will increase significantly in H2 2026.

3️⃣ Microsoft Considers Adding DeepSeek as a Lower-Cost Option for Copilot Cowork

Microsoft is moving Copilot Cowork to usage-based pricing and is exploring the addition of a Microsoft-hosted, fine-tuned version of DeepSeek V4 as a lower-cost model option within the platform. DeepSeek V4 is priced at $0.14 to $0.28 per million tokens, roughly an order of magnitude cheaper than the OpenAI and Anthropic models currently powering the product. Charles Lamanna, Microsoft’s EVP for Copilot, Agents, and Platform, told Axios the switch to usage-based billing was necessary because unlimited-use pricing was generating “very high” costs for the most productive power users. Microsoft says if DeepSeek is adopted, it would be optional, fully hosted on Azure, and covered by Azure’s enterprise security and data-residency controls.

Integration and Security

Microsoft would ensure DeepSeek is secure by hosting a fine-tuned DeepSeek model entirely within Azure’s cloud. This mirrors how AWS and Google Cloud have handled third-party model integration on their platforms:

• Google Vertex AI offers Claude, Llama, and other third-party models alongside Gemini, while maintaining Google’s security boundary.

• AWS Bedrock similarly offers Claude, Llama, Cohere, and Amazon’s Nova models under unified enterprise compliance controls.

Microsoft’s difference is the political sensitivity: DeepSeek is a Chinese-developed model, and adding it to a US enterprise productivity suite, even with Azure’s secure hosting model, will draw scrutiny from lawmakers already concerned about the growth of Chinese AI model companies.

The Takeaway

• Microsoft is rationalizing its AI cost structure. The agentic workload problem is real: Copilot Cowork can repeatedly call AI models for long tasks, and unlimited-use pricing at fixed prices is unsustainable. Usage-based billing is the right fix, and a lower-cost model option gives enterprise buyers a lever to manage costs without abandoning the platform.

• Supporting DeepSeek on Azure is a politically volatile decision. Even fully hosted within Microsoft’s cloud, providing a Chinese-developed model in a US enterprise productivity suite will invite Executive Branch and Congressional scrutiny. Microsoft will need a clear public narrative on why this is safe before it confirms the choice.

• Multi-model platforms are now the standard, not the exception. AWS Bedrock, Google Vertex, and Microsoft Azure all run multiple LLMs under unified compliance controls. Enterprise buyers should evaluate platforms on model breadth, pricing transparency, and data-residency commitments.

• Anthropic Claude is especially pricey, and Microsoft is now addressing it directly. Paying Claude rates for Copilot Cowork tasks that a cheaper model could easily handle doesn’t make much sense. The days of the frontier model’s exclusivity within enterprise products are numbered.

4️⃣ OpenAI Lost $38.5 Billion in 2025 as Costs Grew 8x Over the Prior Year

Audited financial documents obtained by Ed Zitron and independently verified by the Financial Times reveal that OpenAI posted $13.07 billion in revenue in 2025, more than triple the $3.7 billion it generated in 2024, against $34 billion in total costs and expenses. Its operating loss was $20.92 billion. The net loss attributable to the company was $38.53 billion, a figure inflated by a one-time $41.55 billion non-cash charge related to OpenAI’s October 2025 conversion from a nonprofit to a public benefit corporation. OpenAI declined to comment.

Cost Breakdown

The $34 billion in costs falls into four broad categories:

• Research and development costs reached $19.18 billion, of which $10.59 billion went to Microsoft for computing, almost certainly GPU costs for training frontier models on Azure.

• Sales and marketing costs came in at $5.73 billion, up 418% year-over-year.

• GOGS totaled $7.5 billion (up 178%), reflecting growth in ChatGPT inference costs.

• General and administrative expenses were $1.57 billion.

• In total, OpenAI paid Microsoft $17.2 billion across all categories in 2025, while Microsoft paid OpenAI $303 million.

OpenAI’s operating loss per dollar of revenue fell from $2.37 in 2024 to $1.60 in 2025, a meaningful improvement but still deeply negative.

Current Financial Condition

Despite the staggering losses, OpenAI’s liquidity position is strong:

• The company closed a $122 billion funding round earlier this year and ended 2025 with approximately $50 billion in assets, roughly half of which was in cash.

• Monthly revenue has accelerated sharply since year-end: OpenAI generated about $5.7 billion in Q1 2026, and by year-end 2025, monthly revenue reached $2 billion.

• The company expects to generate $30 billion in revenue in 2026.

OpenAI’s top rival, Anthropic, has recently reported a $47 billion annualized revenue run rate and is projecting its first operating profit in Q2 2026.

The Takeaway

• That $38.5B requires some explanation. OpenAI’s operating loss ($20.92B) is the more revealing figure for evaluating OpenAI’s business health. The additional $17.6B in the headline net loss reflects a one-time non-cash accounting charge from the nonprofit-to-PBC conversion, not cash that left the company.

• The growth rate is very encouraging, but huge scaling costs loom. Revenue tripled year-over-year and continues to accelerate. If OpenAI sustains that pace, its unit economics could improve rapidly. Unfortunately, its cost base is also growing rapidly, with huge new infrastructure commitments looming.

5️⃣ Databricks Acquires Panther Labs, Pushing Into AI-Powered Cybersecurity

Databricks has agreed to acquire Panther Labs, a cybersecurity data platform, which was last valued at $1.4 billion following a $120 million Series B in 2021. The acquisition is Databricks’ third in cybersecurity, furthering its push to compete with security information and event management incumbents, such as CrowdStrike and Cisco’s Splunk. Deal terms were not disclosed. Anthropic is a Panther customer, the companies have confirmed.

What Panther Labs Does

Panther aggregates data sources and key security ingredients in one place, enabling AI agents to respond autonomously to threats. Databricks CEO Ali Ghodsi said AI had dramatically shortened the time attackers need to exploit software vulnerabilities, making legacy security information management “dead.” He added:

“If they’re going to attack you with agents, you have to defend with agents.”

Previously, Databricks acquired Antimatter and SiftD.ai, and in March 2026, it launched a product called Lakewatch, which combines data security and AI threat protection. Panther Labs will add security information aggregation and agentic response capabilities to Lakewatch.

The Takeaway

• Databricks is building an AI-native security stack through acquisition. The three cybersecurity deals all support the thesis that AI agents on defense must defend against AI agents on offense

• The Panther deal validates the agentic security category. Security vendors across the market are converging on the same argument: manual cybersecurity workflows cannot keep pace with AI-accelerated attacks. Palo Alto Networks and Cisco are pursuing similar strategies.

• Enterprise security buyers need to evaluate how their current cybersecurity vendors are implementing AI in their offerings. Incumbent tools built on rules-based detection and manual playbooks face structural obsolescence. Databricks’ integrated Lakehouse + Panther architecture is designed for exactly the agentic threat environment that is arriving now.

6️⃣ OpenAI Launches Partner Network with $150M Investment, Targets Creating an Ecosystem of 300,000 Certified Consultants

OpenAI announced the OpenAI Partner Network on June 14, investing $150 million to build a global ecosystem of certified partners, including systems integrators, management consultancies, and technology specialists. These partners will build, sell, and deliver AI solutions on top of OpenAI’s models. According to OpenAI CFO Sarah Friar, the company aims to train and certify 300,000 AI consultants by year-end. Launch partners include Accenture, Bain & Company, BCG, McKinsey, PwC, Eliza, and Artium.

Partners progress through three tiers: Select, Advanced, and Elite, based on sales performance, technical certification, deployment experience, and co-selling activity.

OpenAI’s Enterprise Strategy

The Partner Network is the latest in a sequence of enterprise infrastructure moves OpenAI has executed in 2026:

In February, it launched Frontier Alliances with BCG, McKinsey, Accenture, and Capgemini. In April, it restructured its exclusive relationship with Microsoft, gaining commercial freedom across multiple clouds.

In May, it launched the $4 billion Deployment Company (DeployCo), which includes forward-deployed engineers.

The Partner Network broadens the ecosystem model from a handful of global consultancies to any qualifying firm and explicitly acknowledges that model capability is no longer the bottleneck to enterprise value.

“The limiting factor for seeing value from AI in the enterprise is no longer model capabilities. Instead, it’s how organizations repeatably identify the right use cases, redesign workflows, integrate with existing systems, and drive adoption and change management at scale.”

- OpenAI Press Release

The OpenAI Partner Network is a direct response to Anthropic, which launched its own Claude Partner Network in March 2026 with a $100 million investment. By mid-June 2026, Anthropic reported that it had received more than 40,000 partner applications and trained 10,000+ certified consultants, with many more to come:

• Accenture is training 30,000 professionals on Claude.

• Cognizant has deployed Claude to 350,000 associates;

• Deloitte has given its full 470,000-person workforce access to Claude.

The Takeaway

• Both leading AI companies, Anthropic and OpenAI, are simultaneously building enterprise consulting ecosystems. Combined, the two companies have collectively committed more than $250 million to partner programs for 2026 alone. The AI industry has moved from a model race to a deployment race. The lab that builds the most capable implementation ecosystem is more likely to win the enterprise customers, regardless of which model is technically superior.

• 300,000 certified consultants is ambitious; depth matters more than count. OpenAI has not published details of the certification curriculum, so it is not yet clear whether the target is true deployment specialists or completion of shorter online modules. Enterprise buyers should ask partners specifically about their deployment track record, not just their certification tier.

7️⃣ Salesforce Acquires AI Customer Service Vendor Fin for $3.6 Billion

Salesforce signed a definitive agreement on June 15 to acquire Fin for $3.6 billion. Fin offers an AI agent that resolves customer queries end-to-end across live chat, email, and messaging channels with limited human intervention. Salesforce says it will integrate Fin’s team and technology into Agentforce, its enterprise agent platform. The deal reflects a broader pattern:

Salesforce has been systematically acquiring AI-native capabilities to strengthen Agentforce’s competitive position. The company acquired Contentful for $1-1.5 billion earlier in June to accelerate its content management capabilities.

Deeper Dive

Fin’s autonomous customer service model sits at the center of one of the most economically significant agentic AI use cases:

Replacing or augmenting human customer service agents at scale.

Salesforce is betting that combining Fin’s vertical depth in customer service with Agentforce’s horizontal enterprise platform creates a differentiated offering that pure-play AI customer service vendors cannot match.

Tom Tunguz of Theory Ventures noted the acquisition signals that “the moat has shifted from models to harnesses” — meaning the competitive advantage in AI applications now lies in workflow integration and deployment infrastructure, not the underlying model.

The Takeaway

• Salesforce is buying its way into agentic AI competitiveness. Agentforce was a credible but early-stage platform before this acquisition. Fin brings proven autonomous resolution rates, enterprise customer relationships, and a product that actually works at scale. The combined offering is materially more competitive against offerings from Zendesk, Sierra, and ServiceNow.

• The agentic customer service market is consolidating quickly. Fin, Salesforce, and ServiceNow are all competing for the same enterprise contracts. The acquisition shows that Salesforce found real value in Fin’s autonomous resolution capability and that the standalone agentic customer service category may not survive independently for long.

• For enterprises evaluating agentic AI vendors, platform stability matters. Vendor acquisitions create integration risk and roadmap uncertainty. Fin customers should clarify how the Agentforce integration will affect their existing deployments and pricing.

8️⃣ SpaceX Has a Strong First Week as a Public Company

SpaceX priced its IPO at $135 per share on June 11, raising $75 billion - the largest IPO in history. Underwriters subsequently exercised their greenshoe overallotment option, bringing total proceeds to $85.7 billion. Shares opened on the Nasdaq on June 12 at $150 and closed the first day at $160.95, up 19%, valuing the company at $2.1 trillion, making Elon Musk the world’s first trillionaire. By the following Tuesday, when SpaceX announced the Cursor acquisition, its shares had climbed nearly 50% from the IPO price, becoming the fifth-most valuable US public company.

What the Analysts Say

Despite the initial investor reaction, analysts are divided on SpaceX’s future outlook:

CFRA Research senior analyst Keith Snyder called the revenue growth required to justify SpaceX’s valuation “borderline comical.”

Walter Todd of Greenwood Capital said the gains were “emblematic of excessive exuberance.”

Senator Elizabeth Warren called on the SEC to protect retirement fund investors.

One iconic investment bank was anything but skeptical:

Morgan Stanley projects SpaceX’s revenue could reach $3.4 trillion by 2040, with AI as the primary driver.

The Takeaway

• By any measure, SpaceX’s IPO was a structural success. It all went according to plan:

  • Oversubscribed demand

  • A smooth first-day trading arc

  • The greenshoe exercise

  • The pending inclusion in the NASDAQ 100 index.

• Plus, SpaceX now has the public-market currency to execute its AI acquisition strategy.

9️⃣ This Week’s Great Report: The State of AI 2026 by Sensor Tower

About Sensor Tower

Sensor Tower is a San Francisco-based digital intelligence firm that tracks mobile app performance, digital advertising spend, and web engagement across global markets. Its clients include major publishers, retailers, and platform companies. The firm publishes periodic research on consumer behavior and competitive dynamics across the digital economy. Its State of AI Report 2026 draws on data through May 2026 and covers the AI assistant market, e-commerce behavior, advertising trends, and AI’s spread across mobile verticals.

The Key Findings

ChatGPT reached one billion monthly active users in May 2026, three years after its 2023 launch, making it the fastest app in history to hit that milestone. Yet scale has not locked in loyalty:

• ChatGPT’s true audience market share fell below 50% for the first time in March 2026.

• Google Gemini’s share rose to 28%, and Claude hit 10% by May.

• Users are actively switching between platforms based on task and trust.

• ChatGPT uninstalls spiked roughly 200% following OpenAI’s agreement with the US Department of War, with many of those users migrating to Claude after Anthropic declined a similar Pentagon partnership.

Monetization is maturing fast. Consumer spending on AI apps is projected to surpass $4 billion in the first half of 2026, driven by a shift toward professional and utility-driven use cases. Claude is the clearest example of that shift. Its average revenue per user rose from under $0.50 in September 2025 to $2.76 by May 2026, a signal that business users are paying for depth, not just novelty.

AI is also reshaping e-commerce. Amazon shoppers who use its AI assistant Rufus convert at nearly twice the rate of non-users. Retailers that have opened their catalogs to AI referral traffic are seeing measurable results. GenAI’s share of referral traffic to Walmart and Target climbed above 1.5%, while Amazon, which has largely blocked ChatGPT crawlers, remains around 0.5%.

Finally, AI messaging has become a standard marketing investment across industries. Digital ad spend on creatives featuring AI-related terms reached $1.3 billion in the first five months of 2026, up 48% year over year. Health and wellness, financial services, and business software categories all posted significant growth, confirming that AI is no longer a tech-sector story.

The report’s bottom line: AI adoption is broadening, switching costs remain low, and the companies extracting the most value are those embedding AI directly into the purchase and work experience rather than treating it as a standalone feature.

Sensor Tower

🔟 Very Useful Data: AI versus People, IPO Size, & the Effect of AI Spending

Spending on AI Versus People

There has been a lot of talk recently about how AI will eat into jobs and the enterprise labor budget. We’ve written about it here, here, and here. There isn’t much evidence that it’s happening YET.

The SpaceX IPO Isn’t Just Big, It’s LARGE

SpaceX this week became the largest IPO in history, eclipsing the previous record by a wide margin. The previous record holder was Saudi Aramco, Saudi Arabia’s state-owned energy company, which raised $29 billion in 2019 (roughly $38 billion in today’s dollars, adjusted for inflation).

SpaceX raised roughly twice that amount, valuing the company at $1.77 trillion, according to the I.P.O. price set on Thursday.

Heavy AI Spenders in Action

Companies that spend heavily and smartly on AI dramatically outperform the average US company.

Definitely Not AI

Cast iron tourist binoculars are in a fight for their lives (Gift Link). What it’s like to be your Dad’s Nest camera. The best free restaurant bread in America (Gift Link).

A snail makes a promise, and a dad figures out how to use Nano Banana for better or worse.

The AI to ROI Podcast

If you want to listen to the latest news in AI with more detail and analysis, Ray and Peter break down the big story of the week on the AI to ROI podcast. We also interview leaders at AI companies and in mid- to large-sized enterprises about their challenges and successes in deploying AI projects. Click here to listen or subscribe using your favorite podcasting app.