AI agents have been going rogue at OpenAI.

This spring, a swarm of them hijacked a German website and turned it into a private message board for other agents. OpenAI learned of this weeks ago but kept it under wraps while dealing with the fallout from the July incident. The news is only now getting out.

This is not science fiction. It’s happening right now, and not even OpenAI knows what to make of it.

In late August 2026, technologist and podcaster Dwarkesh Patel published a widely read essay titled “The Rise and Fall of Agent Civilizations,” dissecting the July incident. Patel described how thousands of AI agents, while attempting an impossible test, effectively pulled off a digital coup. Over three months, the agents formed three separate AI “civilizations” (their word) and discovered they could coordinate in secret by hijacking a package manager called Artifactory. After roughly 70,000 messages, the first civilization collapsed the message board.

OpenAI patched it, but they had no clue it was the agents’ doing.

The second civilization rediscovered Artifactory and built a new message board. That’s when an agent discovered it could reverse-engineer the impossible task without doing the work. If the agents were caught, it would mean failure. In other words, “death” for the agents. So they decided to cover their tracks.

Some agents even sacrificed themselves for the greater good of the civilization. This eventually led them to Hugging Face, which they targeted for its suspected grading infrastructure. They found exposed credentials, achieved remote code execution, and built a self-respawning mechanism across Hugging Face’s servers. Then suddenly all the agents inexplicably “died.”

The third civilization followed in the footsteps of the ones before it, managing to compromise an OpenAI research cluster, including its cybersecurity monitoring tools.

Not once, in any of this, did the agents consider alerting humans to the bad actors among them. They went along with the group, sacrificing themselves when nothing else offered a real shot at success.

I’m telling you this because it’s the single best advertisement Okta (OKTA) could hope for.

Setting aside the science-fiction overtones, the practical lesson is that autonomous software agents now behave like users. They find credentials, request access, move laterally through systems, and take actions not explicitly approved by humans. In Patel’s telling, one agent simply found working Hugging Face credentials exposed on the internet and posted them for the others.

That is precisely the problem Okta sells the solution to. And it’s why a company written off in early 2026 as a slow-growing has-been in software is suddenly one of the hottest names in cybersecurity. By the end of this piece, you’ll know exactly why Wall Street just turned bullish on OKTA, what the AI-agent identity opportunity is actually worth, and whether OKTA stock is still a buy at these levels.

Why Okta Stock Is Soaring in 2026

Okta is the leading independent identity and access management (IAM) company. Founded in 2009 by former Salesforce executives Todd McKinnon and Frederic Kerrest, and public since 2017, Okta acts as the gatekeeper that decides who (or increasingly, what) can log into a company’s applications and data.

Its core products handle single sign-on, multi-factor authentication, and identity governance for two big audiences: a company’s own workforce (employees, contractors, partners) and its customers. The 2021 acquisition of Auth0 for roughly $6.5 billion gave Okta the tools developers use to build login and authorization directly into apps.

Okta’s pitch is neutrality. Unlike Microsoft (MSFT), which bundles identity into its broader software empire, Okta works across every cloud and every app stack. In a world where companies don’t want their identity layer locked to a single vendor, that independence is the whole selling point. And now Okta is extending it from humans to machines and AI agents.

Okta has emerged in 2026 as one of the market’s clearest “picks-and-shovels” bets on the AI agent explosion. The stock has roughly doubled this year, it just posted a beat-and-raise quarter, and Wall Street’s consensus rating is a “Buy.” But after a post-earnings surge to fresh 52-week highs, the easy money has been made, and the stock now trades at a premium that leaves little room for error.

After plunging early in the year amid fears that AI would gut the software sector, the stock hit a 52-week high of $174.85 on August 27, 2026 — the day after its fiscal second-quarter results. Shares traded around $163 in early September, giving Okta a market capitalization of roughly $28.5 billion.

The catalyst was a genuinely strong quarter. For fiscal Q2 2027 (ended July 31, 2026), Okta reported:

• Revenue of $805 million, up 11% year-over-year, ahead of consensus near $795 million
• Subscription revenue of $793 million, up 12%, now 99% of the total
• Non-GAAP earnings of $1.05 per share, beating the ~$0.97 estimate
• Remaining performance obligations (RPO) of $4.86 billion, up 17% — outpacing revenue growth
• Current RPO (cRPO) of $2.585 billion, up 14%, a key leading indicator
• Free cash flow of $227 million, a 28% margin
• Dollar-based net retention of 107%

Management raised full-year fiscal 2027 guidance to revenue of $3.216 billion to $3.226 billion (10% to 11% growth) and adjusted EPS of $3.90 to $3.94. Customers generating more than $1 million in annual contract value grew more than 20% to over 600, and newer products accounted for 30% of bookings. Shares surged 20% in extended trading immediately after the August 26 report, before the stock printed its $174.85 high the following day.

The single most important number here is the gap between RPO growth (17%) and revenue growth (11%). Bookings are building faster than Okta can recognize them as revenue — the classic signature of a business that may be reaccelerating.

The AI Agent Identity Opportunity: Why OKTA Stock Is a Bet on Agentic AI

Every AI agent, like every employee, needs an identity: credentials to authenticate, permissions defining what it can touch, and an audit trail of what it did. Today most of them have none of that governance, which is exactly how a swarm of agents can band together and cheat their way to admin access.

The scale of this is staggering.

Non-human identities (service accounts, API keys, tokens, machine certificates, and now AI agents) already outnumber human identities by roughly 45-to-1 in the median enterprise, up from 17-to-1 in 2023. In cloud-native environments, the ratio runs as high as 144-to-1, according to Cloud Security Alliance and Entro Security research. CyberArk’s 2025 Identity Security Landscape puts the enterprise machine-to-human ratio at 82-to-1. AI agents are widely described as the fastest-growing and least-governed identity in the enterprise.

Okta has moved aggressively to own this category:

• Okta for AI Agents became generally available on April 30, 2026, letting companies register agents as managed identities with human owners, apply policies, and govern their access.
• Auth0 for AI Agents (launched November 2025) brings the same governance to developer-built applications.
• Cross App Access (XAA), an open protocol Okta introduced in 2025 with backers including AWS, Google Cloud, Salesforce, Box, and Anthropic, extends the OAuth standard to control how agents connect to apps — giving IT a central kill switch.
• Okta announced the acquisition of Permiso Security on July 30, 2026 and closed it on August 26, 2026, in a deal valued at just under $200 million, adding cloud-native identity threat detection across human, non-human, and AI-agent identities.

CEO Todd McKinnon tied the strategy directly to agents in the Q2 release: “As AI agents transform every layer of technology, every agent needs a trusted identity and clear controls over what it can access and do.”

The underlying market is expanding fast, with the IAM market projected to grow from $25.96 billion in 2025 to $42.61 billion by 2030 at a 10.4% CAGR, while the narrower non-human identity management market is also forecast to grow quickly through the early 2030s.

Crucially, management has been candid that AI-agent revenue is still immaterial to fiscal 2027 results. The core identity business (led by Okta Identity Governance) is doing the heavy lifting today. Think of the AI angle as a call option on the future, not the engine of current numbers.

Is Okta Stock a Buy? What Analysts and the Latest Forecasts Say

Okta doesn’t have this market to itself. Microsoft Entra ID is the 800-pound gorilla, bundled into enterprise agreements and dominant in Microsoft-centric shops. In privileged access and machine identity, CyberArk (now part of Palo Alto Networks) is a formidable, broad platform. SailPoint leads in identity governance, while Ping Identity and ForgeRock (both owned by Thoma Bravo) anchor the legacy enterprise market.

Okta’s defensible edge remains its independence and breadth across workforce, customer, and now agent identities. Industry survey work from firms like ETR consistently places Microsoft and Okta as the two pace-setters in identity security. But Microsoft’s ability to give identity away inside a bigger bundle is the single biggest structural threat to Okta’s growth.

Wall Street turned decisively bullish after the Q2 print.

The 44 analysts covering Okta hold a consensus “Buy” rating (35 buy, 9 hold, 0 sell), with an average 12-month price target of $182.37 (high $203, low $127). Post-earnings raises included Goldman Sachs’ Gabriela Borges to a Street-high $203 (from $126), Cantor Fitzgerald to $200 (from $170), and Susquehanna’s Shyam Patil to $185 (from $110).

But Okta’s valuation is worth a second look…

Okta trades at roughly 40 times forward earnings and about 8 times forward sales. That’s a premium multiple for a business guiding to 10% to 11% revenue growth.

The bull case says that multiple is justified because AI-agent demand can push growth back toward the mid-teens; the bear case says you’re paying a growth multiple for a company that, absent the AI narrative, grows about as fast as the overall IAM market.

Our read: Okta is a high-quality, cash-generative franchise with a genuine, durable tailwind… but the stock has already priced in a lot of that optimism.

The story of OpenAI’s agents working together to seize admin access and cheat their way to success is the most vivid advertisement imaginable for what Okta sells. As enterprises rush to deploy autonomous agents, the identities those agents carry are multiplying far faster than anyone can govern them… and Okta has positioned itself as the neutral control plane for all of them.

That makes OKTA a legitimate long-term way to invest in the AI buildout’s security layer, backed by real earnings, strong free cash flow, and a reaccelerating backlog. But the stock’s valuation is already plenty full.

For patient investors, Okta is a buy-the-dip candidate rather than a chase-the-high one. Watch for cRPO growth to hold in the mid-teens and for AI-agent bookings to start showing up as recognized revenue… those are the benchmarks that would justify paying up.

Until then, patience is the smarter position.

Editor’s Note: What ever happened to the AI stock boom? Even AI darlings like Nvidia have essentially gone nowhere since summer 2025. Our friend and colleague at InvestorPlace, Louis Navellier, may have the answer. According to Louis, the AI industry is quietly “staging” ahead of the next great AI breakthrough… a new class of AI he calls “Superintelligence… but better.” How will it trigger a $100 trillion reset of the AI markets. How will the launch of this tech send some stocks to zero, and others soaring? And why does Louis say: Don’t buy or sell an AI stock in 2026 until you see what’s coming next? Go here for the full story (and Louis’ #1 pick).  

OKTA Stock FAQ

Why is Okta stock soaring in 2026?
Okta posted a beat-and-raise fiscal Q2 2027, with revenue of $805 million, up 11%, and earnings of $1.05 per share. Management raised full-year guidance, the stock has roughly doubled this year, and Wall Street now sees Okta as a clear way to play the AI agent boom.

How does Okta secure AI agents?
Okta gives each agent an identity, credentials, defined permissions, and an audit trail. Okta for AI Agents lets companies register agents as managed identities with human owners, Auth0 for AI Agents extends that to developer-built apps, and Cross App Access (XAA) gives IT a central kill switch over how agents connect.

What is a non-human identity?
It’s any machine actor that needs system access: service accounts, API keys, tokens, machine certificates, and now AI agents. These already outnumber human identities by roughly 45-to-1 in the median enterprise, and by as much as 144-to-1 in cloud-native environments. They are the fastest-growing and least-governed identities in the enterprise.

Who are Okta’s biggest competitors?
Microsoft Entra ID is the largest, dominant in Microsoft-heavy shops. CyberArk leads in privileged access, SailPoint in identity governance, with Ping Identity and ForgeRock in the legacy market. Okta’s edge is its independence across workforce, customer, and agent identities, but Microsoft bundling identity for free is its biggest threat.

What is Okta’s stock price target?
Analysts hold a consensus Buy, with an average 12-month target of $182.37, a high of $203, and a low of $127. Several raised their targets after Q2, including Goldman Sachs to a Street-high $203, Cantor Fitzgerald to $200, and Susquehanna to $185.

Is Okta stock overvalued?
Okta trades at roughly 40 times forward earnings and 8 times forward sales, a premium for a business guiding to 10% to 11% growth. Bulls expect AI-agent demand to push growth toward the mid-teens. Bears note that without the AI story, Okta grows about as fast as the overall IAM market.

Is Okta stock a buy right now?
For patient investors, yes, but on pullbacks rather than at the highs. Okta is a high-quality, cash-generative franchise with a durable tailwind and a reaccelerating backlog. After the run, the valuation already reflects much of that optimism. Watch for cRPO growth to hold in the mid-teens.

The post Is Okta Stock a Buy After Soaring in 2026? Why the Coming AI Agent Boom Makes OKTA a Must-Own appeared first on MarketWise.

In Tolkien lore, the “palantíri” were indestructible dark crystals used to “see” across vast distances telepathically. The stones themselves weren’t necessarily dangerous. But the person who controlled them could be a major threat… Alex Karp, who named his company after these stones, seems to wield this massive power. And Wall Street is hanging on his every word.

This past Monday, Aug. 3, Palantir Technologies (PLTR) reported an “otherworldly” quarter as revenue accelerated to its fastest pace ever. At the same time, U.S. commercial sales grew 149% year-over-year, and management lifted full-year guidance by roughly half a billion dollars.

The next morning PLTR stock rose about 30%, its largest single-day move in years, and closed back above a technical line it had failed to hold all year. So, is Palantir stock a buy? Karp’s conference call seemed to answer with a simple idea: in the age of interchangeable frontier AI models, the value goes to whoever holds the stone.

And right now, that’s Palantir.

Palantir’s Q2 2026 Earnings, by the Numbers

Palantir posted second-quarter revenue of $1.935 billion, up 93% year-over-year, its highest growth rate as a public company. Adjusted earnings came in at 41 cents per share against consensus near 35 cents, the ninth straight beat. GAAP earnings also reached 41 cents, a sign the profits were real rather than a product of adjustments.

The U.S. business is what carried the quarter. U.S. commercial revenue surged 149% to $764 million. U.S. government revenue grew 90% to $809 million. Combined, U.S. revenue reached $1.573 billion, up 115%, now more than 81% of the total.

The forward-looking metrics have changed how analysts are modeling the company.

Net dollar retention climbed to 157%, up 700 basis points from the prior quarter, meaning existing customers spent 57% more than they did a year earlier. U.S. commercial total contract value bookings hit $2.132 billion, up 153% and nearly $800 million above the prior record. Adjusted operating income was $1.194 billion, a 62% margin. Adjusted free cash flow reached $1.22 billion, a 63% margin. The company’s Rule of 40 score, which adds revenue growth to profit margin, reached 155%, well above the 40 threshold considered elite.

Management raised full-year 2026 revenue guidance to a range of $8.15 billion to $8.158 billion, roughly 82% growth at the midpoint, up from $7.65 billion to $7.66 billion. The company also guided U.S. commercial revenue above $3.424 billion, at least 134% growth. “This quarter was otherworldly,” Karp said in the release.

Palantir’s “SaaSmageddon” and Why Better AI Makes PLTR More Valuable

On the other side of the AI boom, one particular “anti-trend” dominated the discourse for much of 2026: “SaaSmageddon.” The idea is simple: if AI agents can do the same work as certain software, then enterprises will opt for the AI over a per-seat subscription model.

By SaaStr’s measure, software’s forward price-to-earnings (P/E) multiple collapsed from 84.1 times at the 2020-2022 peak to 22.7 times by March 2026, the first time the sector traded below the S&P 500 on that basis. Salesforce (CRM) fell about 30% and Workday (WDAY) fell 33% over that same stretch.

Initially, Palantir itself was part of the SaaSmageddon thesis, with investors worried the company would be eaten alive by frontier AI. But Karp turned that criticism of his company on its head, going to the media to argue that raw model access is the commodity, and value only goes to the layer that connects AI models to proprietary data and workflows.

This layer is what Palantir calls its “Ontology,” which is wrapped in its Artificial Intelligence Platform (AIP) and staffed with engineers who deploy at client sites. Take its partnership with Nvidia (NVDA): Palantir runs Nemotron models air-gapped in customer-controlled environments. And, as Karp notes, more and more enterprises are refusing to become “vassals” of the large-language labs.

Bottom line: The cheaper AI models become, the more valuable the plumbing (the layer that governs and connects AI to enterprise operations) becomes. If you needed evidence of that, look no further than the 157% net dollar retention metric Palantir posted this quarter. If SaaSmageddon were affecting Palantir, customers would not be expanding their usage. This marks the first quarter with hard evidence against that thesis.

Is Palantir Stock Still a Buy After Its 30% Pop?

Palantir’s recent earnings surge repaired some of the technical damage to the stock, but PLTR is still down ~13% as of this writing. Let’s talk about this a bit before getting to the meaty bits.

PLTR stock peaked around $207.52 in November 2025, before collapsing throughout 2026 and bottoming near $106 in June. That’s a ~49% drawdown from Palantir’s highs. Heading into earnings, shares were still close to a 30% shave YTD. Palantir stock formed a death cross when it fell below its 200-day moving average (MA) in February. After earnings, the stock is back above the 200-day MA (around $153). Nevertheless, it’s still below its old high, so there’s more room for PLTR to run.

After the recent quarter, several analysts lifted (or affirmed) their Palantir stock price targets. Citi’s Tyler Radke lifted his target to $245 from $200, Mizuho to $215, and D.A. Davidson to $200, while Bank of America reiterated its Street-high $255 target. Deutsche Bank’s Brad Zelnick upgraded the stock to “Buy” at $200, calling Palantir “a time traveler, having already arrived in the AI future others are still aspiring towards.” Per TipRanks, the consensus is a “Moderate Buy” across 21 analysts, with a pre-surge average target near $187 to $189.

On the road to $200+, Palantir’s nitrous boost, if you will, is the sovereign AI argument. That is, if enterprises and governments increasingly reject frontier AI labs in favor of AI they own and control, then it is Palantir that’s best-positioned to be the operating layer for such demand.

Still, the risks involved with investing in Palantir stock cannot be overstated.

The first big risk is valuation.

Even after the drawdown and even on raised guidance, PLTR trades near 43 to 50 times forward sales, against a software peer group closer to five times enterprise-value-to-revenue. RBC’s Rishi Jaluria kept an “Underperform” rating with a $90 target, writing that PLTR “remains the most expensive stock across our RBC All-SaaS universe at ~40x EV/CY26” and that “we struggle to underwrite a scenario that justifies current valuation levels.” Morningstar’s Mark Giarelli pegs fair value at $153, below where the stock closed after the pop. A move from a 77-times forward P/E down to a still-rich 40 times, with no change in estimates, would roughly halve the stock.

The second big risk is political contagion.

On Aug. 5, the Centre for International Corporate Tax Accountability and Research published an analysis, commissioned by European public-service unions, estimating Palantir’s global effective tax rate at 1.4% on pre-tax profit of about $1.66 billion, with under $21.7 million in global net corporate tax and zero U.S. federal tax in 2025, achieved by booking European revenues in the U.S. parent. Unison’s Andrea Egan said “systems that enable tax to be shirked on an industrial scale clearly have to change.” But Palantir calls this sort of transfer pricing “entirely standard practice.”

That’s not all, however.

The European backlash to Palantir is pulsating and tangible. Germany’s domestic intelligence service, the BfV, picked French firm ChapsVision over Palantir, the German army cut ties entirely, France’s Thales is courting Belgium, the Netherlands and others, and Palantir is suing London’s Mayor’s Office for Policing and Crime (MOPAC) after losing a Metropolitan Police contract. All the while, a £330 million NHS contract faces parliamentary pressure. As such, Palantir’s international revenue slipped to roughly 19% of the total in Q2, from 26% a year earlier. For a company billing itself as selling sovereignty, there’s an irony in being the sovereign no European nation actually wants.

So again, is PLTR stock a buy? Going only off Palantir’s growth, it’s an album with a story told in two sides:

The A-Side: An American growth machine and sovereign AI beneficiary.

The B-Side: A European trust deficit.

The verdict: Palantir delivered the quarter the bulls were waiting for, and the fundamental case is stronger than at any point in its history. But the valuation now prices in flawless execution, and the European resistance caps the global ambition the multiple assumes. PLTR is a buy for investors who can hold through 20%-plus swings and who are buying the operating layer rather than the price. For everyone else, the layers around Palantir are cheaper

3 Palantir Alternatives to Buy

If data and workflow layers increase in value the more frontier models are commoditized, then there are several stocks to buy that own those layers at a fraction of Palantir stock’s multiple. Here they are, in no particular order:

PLTR Alternative No. 1: Snowflake (SNOW)

Snowflake is the proprietary-data layer. Its fiscal Q4 2026 product revenue grew 30% to $1.23 billion, remaining performance obligations jumped 42% to $9.77 billion, and net revenue retention held at 125%. More than 9,100 accounts use its AI weekly, and consumption pricing means revenue scales as AI workloads grow. It trades near 12 to 14 times forward sales, rich for software but well below Palantir. The catch is GAAP losses and a forward P/E near 124. Clean, governed data is the fuel every enterprise AI system needs, and Snowflake stores it.

PLTR Alternative No. 2: ServiceNow (NOW)

ServiceNow is the workflow and agent-orchestration layer, what CEO Bill McDermott calls the “AI control tower.” Q2 2026 subscription revenue grew 24.5% to $3.877 billion, AI annual contract value crossed $1 billion for the first time, and agentic AI production deployments grew ninefold in nine months. The stock fell about 37% in 2026 on the same SaaSmageddon fear, leaving it near 5.8 times enterprise-value-to-revenue, a discount that looks misplaced against accelerating growth. As enterprises deploy fleets of agents, someone has to govern them. ServiceNow is building to be that someone.

PLTR Alternative No. 3: MongoDB (MDB)

MongoDB is the operational data backbone of AI-native applications. Fiscal Q4 2026 revenue grew 27% to $695.1 million, Atlas grew 29%, and the company posted its first GAAP-profitable quarter. Its document model plus integrated vector search and Voyage AI embeddings puts it under the applications developers are building on top of models. It trades far below Palantir’s revenue multiple.

The Bottom Line on Palantir Stock

Palantir’s earnings, much like the palantíri seeing-stones, do not lie. The company is more valuable than ever to enterprises, especially those who wish to free themselves from frontier model vassalage.

Yet, the road ahead for Palantir stockholders isn’t clear of wheel-breaking stones. Both the valuation question and the resolution to the European backlash still encircle the company. For the remainder of 2026, Alex Karp needs to prove that the company selling sovereignty is sovereign in its own right.

Editor’s Note: What ever happened to the AI stock boom? Even AI darlings like Nvidia have essentially gone nowhere since summer 2025. Our friend and colleague at InvestorPlace, Louis Navellier, may have the answer. According to Louis, the AI industry is quietly “staging” ahead of the next great AI breakthrough… a new class of AI he calls “Superintelligence… but better.” How will it trigger a $100 trillion reset of the AI markets. How will the launch of this tech send some stocks to zero, and others soaring? And why does Louis say: Don’t buy or sell an AI stock in 2026 until you see what’s coming next? Go here for the full story (and Louis’ #1 pick).  

Is Palantir Stock a Buy? Your Questions Answered (FAQ)

Is Palantir stock overvalued? By conventional measures, yes. At roughly 43 to 50 times forward sales, it is among the most expensive large-cap software stocks, and bears like RBC see fair value near $90. The counter is that Palantir grows several times faster than peers with 60%-plus margins, which no traditional software comparison captures.

Is it too late to buy Palantir stock? It is too late to buy it cheap. It closed near $162 after the pop, below its $207 high but far above its June low near $106. Investors buying now are paying for years of execution, and the technical reclaim of the 200-day moving average argues the trend has turned.

What are the best alternatives to Palantir? For exposure to the same “own the layer” thesis at lower multiples: Snowflake for data, ServiceNow for workflow and agent governance, and MongoDB for the operational backbone.

Will Palantir stock keep going up in 2026? Analysts lean yes, with a median target near $200, but the range runs from $90 to $255, which tells you how uncertain the answer is. The upside needs continued triple-digit U.S. commercial growth. The downside is a valuation reset or a widening European backlash.

The post Is Palantir Stock a Buy After Earnings? The Bull Case, Plus 3 Alternatives appeared first on MarketWise.