The Homeowner Is Fine. The Housing Market Isn’t.

Home Depot’s own framing of its customer has held up so far this year. When the company beat first-quarter expectations in May while reaffirming full-year guidance, chief financial officer Richard McPhail described the core shopper as insulated from the broader consumer slowdown.

“The homeowner in a relative sense is perhaps more protected financially than other customer cohorts and so we continue to see engagement.”
— Richard McPhail, EVP and Chief Financial Officer, The Home Depot, CNBC interview, May 19, 2026

That thesis rests on a homeowner base that is largely locked into pandemic-era mortgages and shielded from monthly rate resets. It says nothing about the transaction economy Home Depot depends on for growth: existing-home sales, which determine how many people are painting, flooring, and renovating a house they just bought. Those sales fell 2.4 percent in June, and the National Association of Realtors has pointed to mortgage-rate sensitivity as the reason buyers keep stepping back in and out of the market.

Q1 Was a Clean Quarter. Q2 Is a Harder Comp.

Home Depot’s first quarter gave management room to sound confident. Total sales rose 4.8 percent to $41.8 billion, comparable sales edged up 0.6 percent, and Pro-focused categories outperformed the DIY side of the business. The company held its full-year targets in place: comp sales of flat to 2 percent, total sales growth of 2.5 to 4.5 percent, and adjusted operating margin of roughly 12.8 to 13 percent.

Wall Street now expects roughly $4.71 in per-share earnings on close to $47.5 billion in revenue for the quarter. That is a step up from the first quarter’s pace, and it comes at a moment when mortgage rates have moved in the wrong direction and existing-home turnover has softened rather than improved. A beat here would say Home Depot’s Pro and SRS businesses can keep growing through a housing market that is not cooperating. A miss would say the resilience story has limits.

Tariffs Are Still Sitting on Top of the Rate Problem

Rates are not the only cost pressure layered into this quarter. The flat 10 percent global import surcharge that had applied since spring expired on July 24, and importers now face a new forced-labor duty on Chinese-origin goods, alongside the Section 301 tariffs on building materials that were never touched by this year’s Supreme Court ruling. Home Depot’s finance team has already demonstrated it can hit a gross-margin target while absorbing tariff headwinds, as it did last year, but the mix of duties it is managing has shifted again since the first quarter closed, a dynamic that has pushed large importers to reconsider how they disclose the exposure this earnings season.

None of that shows up as a single line item investors can isolate. It shows up in how much pricing Home Depot has to pass through without losing the Pro customer who is still spending, and how much of that gets absorbed instead in the margin guidance the company has so far chosen not to touch.

Lowe’s Reports the Next Morning

Home Depot does not report in isolation this cycle. Lowe’s follows a day later, on August 19, giving investors a rare back-to-back read on the same housing-adjacent demand environment from two retailers with different customer mixes. Lowe’s has leaned harder into Pro-segment acquisitions this year, which makes the comparison less like-for-like than it once was, but a divergence between the two prints, one holding guidance and one trimming it, would be the clearest signal yet of whether the sector’s resilience is broad or concentrated in a handful of categories.

Either way, the two reports together will tell a more complete story than either one alone. A market that has priced Home Depot on the assumption that its homeowner is different from everyone else’s consumer is about to find out how much of that difference survives a mortgage rate back near its cycle high.

What to Watch Next

• Whether comp sales land at or above the flat-to-2-percent guidance range, and whether Pro and SRS growth continue to outpace DIY.
  
• Any change to full-year gross margin or operating margin guidance tied specifically to the post-July 24 tariff environment.
  
• Lowe’s August 19 results, and whether its guidance moves in the same direction as Home Depot’s the day before.
  
• Weekly Freddie Mac mortgage-rate prints through the rest of August, since a move back above 6.75 percent would pressure the existing-home turnover Home Depot’s non-Pro business depends on.
  
• July and August existing-home sales data from the National Association of Realtors, the clearest proxy for whether the transaction slowdown is stabilizing or deepening.

The round was led by Norwest, with new investors SoftBank Vision Fund 2, Hitachi Ventures, LG Technology Ventures and Qumra Capital joining alongside existing backers Vertex Ventures, Third Point Ventures, DTCP and Intel Capital. It brings Zenity’s total disclosed funding to roughly $185 million, a jump that maps almost exactly onto how fast enterprises have moved from chatbots that answer questions to agents that take actions.

From Chatbots to Systems With Hands

Zenity, founded in Tel Aviv in 2021 by Ben Kliger and Michael Bargury, started out securing low-code and no-code applications before pivoting to what it now calls AI agent security and governance. The distinction the company draws is specific: a chatbot answers a prompt, but an agent can access enterprise data, invoke software tools, and execute business processes on its own, which creates an attack surface that model-layer or prompt-layer security tools were never built to cover.

That framing is why the company’s research arm has spent the past year disclosing vulnerabilities in agent platforms from major labs rather than just selling a dashboard. It’s also the argument Norwest partner Assaf Harel made in backing the round, saying enterprises need “a new approach to security built for this new and continuously evolving reality” as AI agents get embedded in critical workflows.

“As our use of AI agents continues to grow, maintaining security, governance and operational control is essential. Zenity enables us to confidently deploy AI agents across the enterprise, giving us the visibility and governance required to support innovation at scale.”
— Tadashi Iida, Senior Vice President, CISO and CRO, SoftBank Corp, August 3, 2026

The Money Trail Tells Its Own Story

Zenity’s fundraising history tracks the industry’s own escalation in how seriously it takes agent risk. The company raised $16.5 million in a Series A led by Intel Capital in September 2023, when its pitch was still mostly about low-code application security. By October 2024, a $38 million Series B co-led by Third Point Ventures and DTCP had pushed total funding past $55 million, coinciding with Microsoft’s venture arm M12 making a strategic investment. Less than two years later, the Series C more than tripled the company’s cumulative capital in a single round.

Zenity’s CEO, Ben Kliger, has pointed to the company’s research output as its edge in a field that’s getting crowded fast, saying the research lab “regularly uncovers vulnerabilities and uncontrolled AI-agent behavior inside enterprises” and that the findings feed directly back into the product. That research credibility is also what earned Zenity a place in Gartner’s April 2026 report on AI agent governance vendors, one of the first formal industry assessments of the category.

What the Round Doesn’t Tell You

Kliger declined to disclose Zenity’s post-money valuation, its profitability, or whether an IPO is under consideration, which means the market still has no clean way to size what “AI agent security” is actually worth as a standalone category. That’s a familiar gap for agentic AI more broadly, where enterprise contracts are scaling but the market hasn’t settled on how to price the risk layer sitting underneath them.

What is measurable is the pace. A sector that didn’t really exist as a distinct funding category three years ago just produced a $125 million round from a syndicate that includes a hyperscaler’s own venture arm, a Japanese telecom giant’s strategic fund, and two firms that led its prior rounds returning for a third time. For investors trying to figure out where enterprise AI spend actually lands once the model layer commoditizes, agent governance and security is emerging as one of the clearer answers.

What to Watch Next

• Whether Zenity discloses a post-money valuation or IPO intentions as the round closes and the company scales its go-to-market team.
  
• Additional AI agent security funding rounds from competitors, which would confirm this is a category-wide repricing rather than a single-company outlier.
  
• Any formal enterprise disclosure of an AI agent security incident at a Fortune 500 company, which would be the clearest validation yet of the risk Zenity is underwriting against.
  
• Whether SoftBank Vision Fund 2’s participation signals a broader push into AI security infrastructure investments following its existing AI portfolio commitments.
  
• Gartner’s next AI agent governance vendor assessment, expected to show whether Zenity’s category leadership is holding as rivals raise their own rounds.

Buried inside the Commission’s Spring 2026 Unified Agenda of Regulatory and Deregulatory Actions is a listing that matters more than its bureaucratic name suggests: a proposal to update the definition of an accredited investor and simplify what the agenda calls exempt offering pathways. Chairman Paul Atkins framed it as part of a broader push to modernize capital formation. For anyone tracking the crowdfunding platforms, secondary markets, and pre-IPO vehicles this publication covers, it is the clearest signal yet that the SEC intends to move on its own timeline, independent of whatever Congress eventually does with the INVEST Act.

The Agenda Says the Quiet Part Out Loud

The Spring 2026 agenda runs to roughly three dozen proposed rulemakings, but the private markets section is where Atkins spent his own statement’s attention. He described a mandate to make exposure to private markets available beyond a narrow band of wealthy insiders, while preserving investor protection guardrails.

“Should not be reserved for wealthy insiders.”
— Paul S. Atkins, Chairman, U.S. Securities and Exchange Commission, July 7, 2026

That is a notable thing for a sitting SEC chairman to say about the wealth and income thresholds his own agency enforces. It also lines up with where Atkins has already put his agency’s weight. In April, the Commission asked a federal court to dismiss a lawsuit challenging the current accredited investor income test, even as commissioners including Atkins have separately acknowledged in public remarks that the wealth-and-income test does not actually measure financial sophistication.

Congress Has the Bill. The SEC Has the Calendar.

The INVEST Act, which would direct the SEC to add knowledge and experience-based pathways to accredited status and raise the Regulation A+ offering cap from $75 million to $150 million, passed the House on a lopsided 302-123 vote back in December. Since then it has mostly sat with the Senate Banking Committee, which has been consumed by crypto market structure negotiations and housing legislation.

House Republicans are now pushing the Senate to act before the current Congress runs out the clock. Reporting from Punchbowl News quotes Representative Ann Wagner, who chairs the House Financial Services Subcommittee on Capital Markets, saying “what the Senate doesn’t have is time” as lawmakers weigh whether the bill can move before a lame-duck session swallows the calendar.

That is the contrast worth sitting with. A bipartisan House bill with a specific number attached to it, an expanded Reg A+ cap, a defined accredited investor test, has been parked in a Senate committee for months. An executive agency, acting under authority it already has, put the same basic goal on its own rulemaking calendar in a single afternoon.

Why the Sequencing Matters More Than the Substance

If the SEC proposes and finalizes its own accredited investor rule before the Senate moves, large portions of the INVEST Act’s private markets provisions become redundant, and the legislative fight narrows to whatever only Congress can do, such as statutory Reg A+ cap changes or 403(b) plan parity. If the Senate acts first, it hands the SEC a mandate rather than the other way around. Either path changes the addressable market for the platforms that depend on exemption thresholds.

It is worth being precise about what has and has not happened. The Spring 2026 Unified Agenda is a list of intentions with rough timeframes the SEC itself describes as guidelines rather than commitments. No proposed rule text has been published, no comment period has opened, and the scope of the “exempt offering pathways” language remains vague even to securities lawyers who track the agenda closely. This is a placeholder, not a rule.

What makes it worth tracking now, rather than waiting for a formal proposal, is the combination of signals arriving at once: a chairman’s public statement, a court filing defending the agency’s authority to act unilaterally, and an agenda item that survived from the prior year’s version. None of those are binding. Together they describe an agency that has decided the accredited investor definition is worth spending political capital on, with or without the Senate.

The Platforms Watching the Clock

Every crowdfunding and secondary-market platform this publication tracks has some exposure to where the accredited line ends up. A broader definition expands the pool of investors platforms like Republic and StartEngine can market accredited-only products to, potentially undercutting some of the appeal of Reg CF’s non-accredited access. A narrower Reg A+ cap fight, meanwhile, stays entirely in Congress’s hands regardless of what the SEC does on the definition itself. The two tracks are related but not identical, and issuers building toward a 2027 capital raise are effectively betting on which one moves first.

For now, the SEC has the more concrete near-term catalyst. A proposed rule, even an unpolished one, would give the market something to comment on and a timeline to hold the agency to. The Senate has neither, and December is closer than it looks on a committee calendar that has not touched the bill since referral.

What to Watch Next

• Whether the SEC publishes an actual Notice of Proposed Rulemaking on the accredited investor definition, rather than leaving it as an agenda placeholder, and what specific pathways (exam-based, credential-based, income-indexed) the proposal contains.
  
• Whether the Senate Banking Committee schedules any markup of the INVEST Act before the end of the year, which would determine whether Congress or the SEC sets the terms first.
  
• Whether Chair Atkins’ agency moves independently on Regulation A+ or Regulation CF caps, which the current agenda language does not explicitly address, separate from what Congress does with the $150 million Reg A+ figure in the INVEST Act.
  
• How platforms including StartEngine, Republic, and Wefunder position their product mix in public commentary if a formal comment period opens, since their current business models are built around the existing thresholds.
  
• Any court ruling on the pending accredited investor lawsuit the SEC has moved to dismiss, since a ruling against the agency could force the rulemaking timeline to move faster than the agenda currently suggests.

That would be a rough week on its own. It happened four weeks after a separate selloff already cost Alphabet an estimated $225 billion, and it landed in the same month that a Chinese lab released an open-weight model that undercuts Google’s pricing and, on at least one benchmark, beats every closed model on the market except two. None of that shows up in Alphabet’s income statement. All of it shows up in the stock.

The promise that didn’t ship

Sundar Pichai unveiled Gemini 3.5 at Google I/O on May 19, but only the smaller Flash version actually launched that day. The heavier Pro model, the one meant to compete directly with Anthropic and OpenAI’s top-tier systems, was described as still being tested internally.

“We’re also excited for Gemini 3.5 Pro. We are using it internally, it’s showing great improvements, and it will be coming next month.”
— Sundar Pichai, CEO, Alphabet and Google, Google I/O keynote, May 19, 2026

Next month arrived and Pro still hadn’t shipped. According to Bloomberg’s reporting, Google went back and retrained the model on updated data specifically to fix its coding performance, and the results still fell short of what the company was aiming for. There is still no confirmed release date.

This isn’t Google’s first stumble of the summer. In late June, two senior Google DeepMind figures, Gemini co-lead Noam Shazeer and AlphaFold creator John Jumper, announced departures for OpenAI and Anthropic within days of each other. Shares fell as much as 7 percent intraday and closed down roughly 5 percent, wiping out an estimated $225 billion in market value on fears that Google was losing the talent race even as it poured close to $190 billion into AI infrastructure this year.

Add the two events together and Alphabet has shed somewhere in the neighborhood of $425 billion in market value in six weeks, without a single change to reported revenue, margins, or guidance. That is the part worth sitting with. Investors aren’t repricing Alphabet’s business. They’re repricing its ability to keep shipping at the pace the AI race now demands.

Meanwhile, a cheaper model just showed up uninvited

On the same day Bloomberg’s report landed, Beijing-based Moonshot AI released Kimi K3, a 2.8 trillion parameter open-weight model the company describes as the largest ever built. It arrived with a full benchmark table, and independent trackers have started to confirm at least parts of it: Artificial Analysis places K3 fourth on its intelligence index, behind Claude Fable 5 and GPT-5.6 Sol but ahead of Claude Opus 4.8, and Arena’s blind developer testing ranked it first on frontend code generation, ahead of Fable 5.

The pricing is the more uncomfortable detail for incumbents. Moonshot is charging $3 per million input tokens and $15 per million output tokens, roughly half of what Anthropic charges for Opus 4.8, and the full model weights are due for public release by July 27. Moonshot itself raised $2 billion at a $20 billion valuation in May and is reportedly now in talks for a round that would value it at $30 billion.

Google’s own Gemini 3.5 Flash, the model that did ship in May, already beat its own previous-generation flagship on several agentic benchmarks. That comparison mattered when Flash was the newest thing on the market. It matters less now that Kimi K3 is undercutting both Flash and Pro on price while landing in the same performance tier as Google’s still-unshipped model.

Wednesday is the real test

Alphabet reports second-quarter earnings on July 22, and the timing could not be more pointed. The company heads into the print with cloud and advertising fundamentals that most analysts still describe as strong, but a debate over AI execution that has now cost more in market value over six weeks than the entire disclosed cost of the Gemini delay itself.

A clean quarter with confident language on Gemini 3.5 Pro’s actual ship date would go a long way toward reversing the narrative. Another vague answer, paired with a rival model from a two-year-old Chinese lab undercutting Google on both price and select benchmarks, would harden a story that has so far been driven entirely by sentiment rather than numbers.

What to Watch Next

• Alphabet’s Q2 2026 earnings call on July 22 for any confirmed Gemini 3.5 Pro launch date, and whether management addresses the coding-performance gap directly.
  
• Independent, third-party verification of Kimi K3’s self-reported benchmarks as more evaluators publish results beyond Moonshot’s own launch data.
  
• Moonshot AI’s full open-weight release, scheduled by July 27, and whether enterprise coding tools begin integrating K3 as a default or fallback model.
  
• Any confirmed Gemini 3.5 Pro launch window from Google, and how its benchmark profile compares to Kimi K3 and GPT-5.6 Sol once it actually ships.
  
• Whether OpenAI or Meta respond to Kimi K3’s pricing with their own cuts, given enterprises are already evaluating models on cost per token as much as raw capability.

This is the quarter that either confirms UnitedHealth fixed the problem or shows the first quarter’s improvement was a one-off.

The Number That Broke, Then Un-Broke

The medical care ratio, the share of premium revenue spent on medical costs, is the single metric that defines whether a health insurer’s pricing is keeping pace with how much care its members actually use. UnitedHealth’s ratio climbed every quarter through 2025: 84.8% in Q1, 89.4% in Q2, 89.9% in Q3, and 92.4% in Q4, a level the company’s own Q4 earnings call described as reflecting elevated utilization alongside a charge tied to cyberattack costs, divestitures, and workforce reductions. The full-year 2025 ratio landed at 89.1%, still the worst annual reading in the company’s recent history.

Then Q1 2026 landed at 83.9%, below every quarter of 2025 and even below the year-ago Q1 2025 print of 84.8%. UnitedHealth attributed the improvement to repricing across its business and favorable reserve development, even as it acknowledged utilization and unit costs remained elevated. The company beat consensus on that print by a wide enough margin that management raised full-year adjusted EPS guidance to above $18.25, up from an earlier target of $17.75.

One clean quarter after four consecutive quarters of deterioration is not the same as a structural fix. The question for July 16 is whether the second quarter’s ratio holds anywhere near Q1’s level, or whether the improvement was front-loaded through favorable reserve releases that won’t repeat.

“We are continuing to help simplify and modernize health care for the people and care providers we serve, bringing greater value, affordability, transparency and connectivity.”
— Stephen Hemsley, Chief Executive Officer, UnitedHealth Group, April 21, 2026

Why the Second Quarter Carries More Weight Than the First

UnitedHealth’s own guidance structure puts extra pressure on this print. Management has signaled that roughly two-thirds of full-year adjusted earnings are expected in the first half, which implies a sequential step-down in EPS from Q1’s $7.23 even in a clean quarter. That framing gives the company some room to miss the headline number without alarming investors, but it also means analysts will be reading through the number rather than at it.

The specific lines to watch inside the print: whether UnitedHealthcare’s operating margin, which improved to 6.6% in Q1 from 6.2% a year earlier, holds as membership continues to contract in Medicare Advantage and ACA plans. Optum Health’s earnings are expected to moderate through the year on seasonality, and the segment is also absorbing the shift of Optum Financial into Optum Insight, which complicates a clean quarter-over-quarter comparison.

UnitedHealth has also resumed share buybacks earlier than planned, targeting at least $2 billion deployed by the end of the second quarter. That is either a signal of management’s confidence in the recovery or a lever being pulled to support the stock while the underlying cost trends are still being proven out. The July 16 print is the first real chance to tell which.

The Sector Is Watching, Not Just the Stock

UnitedHealth’s stock trades at a premium multiple relative to where it sat through 2025, and that premium is built almost entirely on the belief that the medical cost spiral is over. A soft print, or commentary suggesting the Q1 improvement leaned on reserve releases rather than durable repricing, would ripple through the rest of managed care, given how closely peers like Elevance are navigating the same Medicare Advantage cost environment this earnings season.

There is also a governance dimension worth noting heading into this print. UnitedHealth combined its CEO and Chairman roles under Hemsley this year, a structural change that concentrates accountability for the recovery in one person rather than splitting it across a board chair and an operating chief. If the numbers hold, that concentration reads as decisive leadership. If they don’t, there is no obvious hand-off of blame.

What to Watch Next

• The Q2 2026 medical care ratio itself, and whether management characterizes any improvement as durable repricing or as favorable reserve development that won’t recur.
  
• Any revision to full-year adjusted EPS guidance above the current $18.25 floor, which would signal confidence extending beyond a single clean quarter.
  
• UnitedHealthcare membership trends in Medicare Advantage and ACA plans, where continued contraction would show the margin recovery is coming at the cost of market share.
  
• Elevance’s own Q2 print this earnings season, as a read on whether the cost environment is improving industry-wide or UnitedHealth-specific.
  
• Pace of the $2 billion buyback target and whether management extends repurchase activity into the back half of the year.

The first time was Anthropic. On June 12, the Commerce Department issued an export control order that forced Anthropic to take Claude Fable 5 and Mythos 5 offline worldwide, three days after their launch. The controls were lifted June 30, and Anthropic restored access on July 1, with Mythos still limited to a short list of trusted organizations under a program called Project Glasswing. Two labs, two models, the same sequence: ship, get flagged, get gated, get cleared.

A voluntary process that doesn’t feel voluntary

The mechanism behind both episodes is officially optional. President Trump signed an executive order this spring letting developers submit “covered frontier models” to the government for review before release, and OpenAI’s delay ran through the same informal channel: the Office of the National Cyber Director and the Office of Science and Technology Policy asked the company to hold back GPT-5.6 Sol, its top-tier model, citing its cybersecurity capabilities.

OpenAI complied, and Sam Altman made clear he did not love it.

“We’ve made clear to the U.S. government that this is not our preferred long-term model, and will work with them and others in industry to achieve a more sustainable approach for future releases.”
— Sam Altman, CEO, OpenAI, internal memo, June 25, 2026

The Commerce Department’s Center for AI Standards and Innovation ran additional testing during the gated window, and GPT-5.6 Sol scored 96.7 percent on OpenAI’s internal cyberattack evaluation, with ExploitBench results the company says are comparable to Anthropic’s Mythos Preview at roughly a third of the inference cost. That is the exact capability profile the government keeps citing as its reason to slow releases down.

The next 22 days set the template

What happened with Fable, Mythos, and now GPT-5.6 was improvised, negotiated model by model. That changes, or is supposed to, by August 1. The executive order gives the NSA sixty days from signing to finalize a classified benchmarking process for designating covered frontier models, and requires a multi-agency group to publish a formal voluntary framework for how the review actually works.

Separately, Anthropic, OpenAI, Google, Microsoft, and Amazon have been in talks reported in early July toward a shared jailbreak severity scoring system, modeled loosely on how the cybersecurity industry already grades vulnerabilities. If that lands alongside the August 1 deadline, frontier labs get something they don’t have today: a predictable release calendar instead of a case-by-case negotiation with whichever officials happen to be in the room.

Why this matters more than the benchmark numbers

For investors modeling OpenAI and Anthropic ahead of their expected public listings, model-release timing has quietly become a disclosed risk factor rather than an internal scheduling detail. A delay tied to national security review is a different kind of delay than a compute shortage or a training run that ran long, and it is one neither company fully controls. OpenAI is separately weighing a 5% stake for the U.S. government ahead of its IPO, a proposal that reads differently once you’ve watched Washington demonstrate it can already gate a launch without owning a share of the company.

The competitive read is straightforward enough. Whichever lab experiences the shorter, more predictable review window has a real advantage heading into a roadshow, because investors will price in execution risk on future model releases the same way they price in any other supply chain dependency. A twelve-day gate that ends in a clean, broad launch is a very different data point than a suspension that goes on for weeks with no fixed end date.

The open question is who writes the rules

Both companies say they want a repeatable process. Neither has said what happens if the August 1 framework arrives and a lab disagrees with where the government draws the line on the next model. Until that framework exists in writing, every frontier release from every major lab carries the same asterisk: cleared by the developer, subject to change by whoever is reviewing it in Washington that week.

What to Watch Next

• Whether the NSA and the multi-agency group meet the August 1 deadline for a formal, published voluntary framework, or whether the process remains as ad hoc as it was for Fable, Mythos, and GPT-5.6.
  
• Progress on the five-lab jailbreak severity scoring talks involving Anthropic, OpenAI, Google, Microsoft, and Amazon, and whether a shared standard actually shortens future review windows.
  
• Whether Mythos Preview’s restricted “trusted organizations” access expands, or whether GPT-5.6 Sol’s broader public launch becomes the model other labs are asked to match.
  
• How OpenAI’s discussions over a government equity stake evolve alongside this review process, and whether a financial relationship changes the terms of future model clearances.