Runpod, an AI app hosting platform that launched four years ago, has hit a $120 million annual revenue run rate, founders Zhen Lu and Pardeep Singh tell TechCrunch.  

Their startup journey is a wild example of how if you build it well and the timing is lucky, they will definitely come.

The story includes bootstrapping their way to over $1 million in revenue; landing a $20 million seed round after VC Radhika Malik, a partner at Dell Technologies Capital, saw some Reddit posts; and gaining another key angel investor, Hugging Face co-founder Julien Chaumond, because he was using the product and reached out over the support chat, the founders tell TechCrunch. 

It all began in late 2021 when the two friends, who worked together as corporate developers for Comcast, decided the hobby they were doing wasn’t fun anymore. 

They had built setups of specialized computers used to generate Ethereum in their respective New Jersey basements. While they did successfully mine a bit of the cryptocurrency, it wasn’t enough to pay back their investment, they said. Plus, mining was going to end after the much-ballyhooed network upgrade called “The Merge.” 

On top of that, it was “boring” after a couple of months, Lu said. 

But they had talked their wives into letting them spend a good $50,000 on the hobby between them, they estimated. Lu and Singh knew that home harmony depended on finding a way to use those GPUs. 

The devs had been engaged in machine learning projects at work, so they opted to convert their mining rigs into AI servers. This was before ChatGPT, even before DALL-E 2. 

As they repurposed the rigs, “We were seeing how really god-awful the software stack was for dealing with these GPUs,” Lu said. As developers, they found a problem they wanted to solve.  

Runpod was born “because we felt that the actual experience of developing software on top of GPUs was just hot garbage,” Lu described. 

A few months later in early 2022, they were ready to share what they had built. Runpod is a platform for hosting AI apps, emphasizing speed, easily configured hardware (including a serverless option that automates configuration), and dev tools like APIs, command-line interfaces, and other integrations.  

Back in 2021, they only had a few such integrations (like support for popular web app tool Jupyter notebooks). The next problem: finding beta testers. 

“As first-time founders, we didn’t really know how to market or how to do anything,” Lu recalled. “So I’m like, all right, let’s just post on Reddit.” 

So, they posted in a couple of AI-oriented subreddits. The offer was simple: free access to their AI servers in exchange for feedback. It worked. They landed beta customers, which led to paying customers. Within nine months, they had quit their jobs and hit $1 million in revenue, they said. 

Bootstrapping growth

But that led to another problem. “Six months in, business users were like, ‘Hey, I want to actually run real business stuff on your platform. But I cannot run it on servers that are in people’s basements,” Lu said. 

It had not occurred to the New Jersey founders to raise capital from VCs. Instead they formed revenue-share partnerships with data centers to grow capacity. But it was stressful. The founders needed to stay three steps ahead. 

“If we don’t have the GPUs, the market sentiment, the user sentiment changes. Because when they don’t see capacity from you, they go somewhere else,” Singh described. 

Meanwhile, their user base was growing on Reddit and Discord, especially after ChatGPT launched.

VCs were also on the prowl for investments. Malik saw them on Reddit and reached out, their first VC call. But Lu didn’t know how to pitch to an investor. “Radhika was super helpful, even at the first conversation,” he said. She basically explained to him how a VC thinks and told him she’d stay in touch.  

Meanwhile, Lu had a business to run that had to pay for itself. “It was almost two years where we really didn’t have any funding,” he said. So Runpod never offered a free tier. It had to at least pay for itself, even if it wasn’t throwing off much profit. Unlike other AI cloud services that began as crypto miners, these founders refused to take on debt, they said. 

By May 2024, with AI app fever spreading, their lucky decision to launch AI hosting for devs two years earlier was paying off. Their business had grown to 100,000 developers, and they landed a $20 million seed deal co-led by the VC arms of both Dell and Intel, with participation from big names like Nat Friedman and Chaumond.  

They haven’t raised more money since but are now planning to, armed with a business that, they believe, should command a healthy Series A. 

Today, Runpod counts 500,000 developers as customers, ranging from individuals to Fortune 500 enterprise teams with multimillion-dollar annual spend, the founders said. 

Their cloud spans 31 regions globally and counts customers like Replit, Cursor, OpenAI, Perplexity, Wix, and Zillow as users.

Competition is also fierce. Devs have all the major clouds to choose from (AWS, Microsoft, Google), plus plenty of industry-specific choices like CoreWeave and Core Scientific. 

But they also see their place in the world a bit differently — as a dev-centric platform. They don’t see coding ever going away but changing. Programmers will become AI agent creators and operators. 

“Our goal is to be what this next generation of software developers grows up on,” Lu said.  

On Tuesday, U.K.-based Iranian activist Nariman Gharib tweeted redacted screenshots of a phishing link sent to him via a WhatsApp message.

“Do not click on suspicious links,” Gharib warned. The activist, who is following the digital side of the Iranian protests from afar, said the campaign targeted people involved in Iran-related activities, such as himself.

This hacking campaign comes as Iran grapples with the longest nationwide internet shutdown in its history, as anti-government protests — and violent crackdowns — rage across the country. Given that Iran and its closest adversaries are highly active in the offensive cyberspace (read: hacking people), we wanted to learn more. 

Gharib shared the full phishing link with TechCrunch soon after his post, allowing us to capture a copy of the source code of the phishing web page used in the attack. He also shared a write-up of his findings.

TechCrunch analyzed the source code of the phishing page, and with added input from security researchers, we believe the campaign aimed to steal Gmail and other online credentials, compromise WhatsApp accounts, and conduct surveillance by stealing location data, photos, and audio recordings. 

It is unclear, however, if the hackers were government-linked agents, spies, or cybercriminals — or all three. 

TechCrunch also identified a way to view a real-time copy of all the victims’ responses saved on the attacker’s server, which was left exposed and accessible without a password. This data revealed dozens of victims who had unwittingly entered their credentials into the phishing site and were subsequently likely hacked.

The list includes a Middle Eastern academic working in national security studies; the boss of an Israeli drone maker; a senior Lebanese cabinet minister; at least one journalist; and people in the United States or with U.S. phone numbers. 

TechCrunch is publishing our findings after validating much of Gharib’s report. The phishing site is now down.

Inside the attack chain

According to Gharib, the WhatsApp message he received contained a suspicious link, which loaded a phishing site in the victim’s browser.

The link shows that the attackers relied on a dynamic DNS provider called DuckDNS for their phishing campaign. Dynamic DNS providers allow people to connect easy-to-remember web addresses — in this case, a duckdns.org subdomain — to a server where its IP address might frequently change. 

It’s not clear whether the attackers shut down the phishing site of their own accord or were caught and cut off by DuckDNS. We reached out to DuckDNS with inquiries, but its owner Richard Harper requested that we send an abuse report instead.

From what we understand, the attackers used DuckDNS to mask the real location of the phishing page, presumably to make it look like a genuine WhatsApp link. 

The phishing page was actually hosted at alex-fabow.online, a domain that was first registered in early November 2025. This domain has several other, related domains hosted on the same dedicated server, and these domain names follow a pattern that suggests the campaign also targeted other providers of virtual meeting rooms, like meet-safe.online and whats-login.online.

We’re not sure what happens while the DuckDNS link loads in the victim’s browser, or how the link determines which specific phishing page to load. It may be that the DuckDNS link redirects the target to a specific phishing page based on information it gleans from the user’s device.

The phishing page would not load in our web browser, preventing us from directly interacting with it. Reading the source code of the page, however, allowed us to better understand how the attack worked.

Gmail credential and phone number phishing

Depending on the target, tapping on a phishing link would open a fake Gmail login page, or ask for their phone number, and begin an attack flow aimed at stealing their password and two-factor authentication code. 

But the source code of the phishing page code had at least one flaw: TechCrunch found that by modifying the phishing page’s URL in our web browser, we could view a file on the attacker’s servers that was storing records of every victim who had entered their credentials. 

The file contained over 850 records of information submitted by victims during the attack flow. These records detailed each part of the phishing flow that the victim was in. This included copies of the usernames and passwords that victims had entered on the phishing page, as well as incorrect entries and their two-factor codes, effectively serving as a keylogger. 

The records also contained each victim’s user agent, a string of text that identifies the operating system and browser versions used to view websites. This data shows that the campaign was designed to target Windows, macOS, iPhone, and Android users.

The exposed file allowed us to follow the attack flow step-by-step for each victim. In one case, the exposed file shows a victim clicking on a malicious link, which opened a page that looked like a Gmail sign-in window. The log shows the victim entering their email credentials several times until they enter the correct password. 

The records show the same victim entering their two-factor authentication code sent to them by text message. We can tell this because Google sends two-factor codes in a specific format (usually G-xxxxxx, featuring a six-digit numerical code).

WhatsApp hijack and browser data exfiltration

Beyond credential theft, this campaign also seemed to enable surveillance by tricking victims into sharing their location, audio, and pictures from their device.

In Gharib’s case, tapping on the link in the phishing message opened a fake WhatsApp-themed page in his browser, which displayed a QR code. The lure aims to trick the target into scanning the code on their device, purportedly to access a virtual meeting room.

Gharib said the QR code was generated by the attacker, and scanning or tapping it would instantly link the victim’s WhatsApp account to a device controlled by the attacker, granting them access to the victim’s data. This is a long-known attack technique that abuses the WhatsApp device linking feature and has been similarly abused to target users of messaging app Signal.

We asked Granitt founder Runa Sandvik, a security researcher who works to help secure at-risk individuals, to examine a copy of the phishing page code and see how it functions. 

Sandvik found that when the page loaded, the code would trigger a browser notification asking the user for permission to access their location (via navigator.geolocation), as well as photos and audio (navigator.getUserMedia). 

If accepted, the browser would immediately send the person’s coordinates to the attacker, capable of identifying the location of the victim. The page would then continue to share the victim’s location data every few seconds, for as long as the page remained open. 

The code also allowed the attackers to record bursts of audio and snap photos every three to five seconds using the device camera. However, we did not see any location data, audio, or images that had been collected on the server.

Thoughts on victims, timing, and attribution

We do not know who is behind this campaign. What is clear is that the campaign was successful in stealing credentials from victims, and it is possible that the phishing campaign could resurface. 

Despite knowing the identities of some of the people in this cluster of victims who were targeted, we don’t have enough information to understand the nature of the campaign. The number of victims hacked by this campaign (that we know of) is fairly low — fewer than 50 individuals — and affects seemingly ordinary people across the Kurdish community, as well as academics, government officials, business leaders, and other senior figures across the broader Iranian diaspora and Middle East.

It may be that there are far more victims than we are aware of, which could help us understand who was targeted and potentially why.

The case that this could be a government-backed actor

It is unclear what motivated the hackers to steal people’s credentials and hijack their WhatsApp accounts, which could also help identify who is behind this hacking campaign.

A government-backed group, for example, might want to steal the email password and two-factor codes of a high-value target, like a politician or journalist, so they can download private and confidential information.

That could make sense since Iran is currently almost entirely cut off from the outside world, and getting information in or out of the country presents a challenge. Both the Iranian government, or a foreign government with interests in Iran’s affairs, could plausibly want to know who influential Iranian-linked individuals are communicating with, and what about.

As such, the timing of this phishing campaign and who it appears to be targeting could point to an espionage campaign aimed at trying to collect information about a narrow list of people.

We asked Gary Miller, a security researcher at Citizen Lab and mobile espionage expert, to also review the phishing code and some of the exposed data from the attacker’s server. 

Miller said the attack “certainly [had] the hallmarks of an IRGC-linked spearphishing campaign,” referring to highly targeted email hacks carried out by Iran’s Islamic Revolutionary Guard Corps (IRGC), a faction of Iran’s military known for carrying out cyberattacks. Miller pointed to a mix of indications, including the international scope of victim targeting, credential theft, the abuse of popular messaging platforms like WhatsApp, and social engineering techniques used in the phishing link.

The case that this might be a financially motivated actor

On the other hand, a financially motivated hacker could use the same stolen Gmail password and two-factor code of another high-value target, such as a company executive, to steal proprietary and sensitive business information from their inbox. The hacker could also forcibly reset passwords of their victim’s cryptocurrency and bank accounts to empty their wallets.

The campaign’s focus on accessing a victim’s location and device media, however, is unusual for a financially motivated actor, who might have little use for pictures and audio recordings.

We asked Ian Campbell, a threat researcher at DomainTools, which helps analyze public internet records, to look at the domain names used in the campaign to help understand when they were first set up, and if these domains were connected to any other previously known or identified infrastructure. 

Campbell found that while the campaign targeted victims in the midst of Iran’s ongoing nationwide protests, its infrastructure had been set up weeks ago. He added that most of the domains connected to this campaign were registered in early November 2025, and one related domain was created months back in August 2025. Campbell described the domains as medium to high risk and said they appear to be linked to a cybercrime operation driven by financial motivations.

An additional wrinkle is that Iran’s government has been known to outsource cyberattacks to criminal hacking groups, presumably to shield its involvement in hacking operations against its citizens. The U.S. Treasury has sanctioned Iranian companies in the past for acting as fronts for Iran’s IRGC and conducting cyberattacks, such as launching targeted phishing and social engineering attacks. 

As Miller notes, “This drives home the point that clicking on unsolicited WhatsApp links, no matter how convincing, is a high-risk, unsafe practice.”

To securely contact this reporter, you can reach out using Signal via the username: zackwhittaker.1337

Lorenzo Franceschi-Bicchierai contributed reporting.

Anthropic has appointed Irina Ghose, a former Microsoft India managing director, to lead its India business as the U.S. AI startup prepares to open an office in Bengaluru. The move underscores how India is becoming a key battleground for AI companies looking to expand beyond the U.S. for major growth markets.

Ghose brings deep big-tech operating experience to the role. She spent 24 years at Microsoft before stepping down in December 2025. Her appointment gives Anthropic a seasoned executive with local enterprise and government relationships as it gears up to establish an on-the-ground presence in one of the world’s fastest-growing AI markets.

India has become one of Anthropic’s most strategically important markets, with the country already ranking as the second-largest user base for Claude and usage heavily skewing toward technical and work-related tasks, including software development. Arch-rival OpenAI is also sharpening its focus on the market with plans to open an office in New Delhi — a sign India is fast becoming one of the most contested arenas in the global race to commercialize generative AI.

While India offers enormous scale — with more than a billion internet subscribers and over 700 million smartphone users — converting that reach into meaningful revenue has proven difficult, pushing AI companies to experiment with aggressive pricing and promotions. OpenAI last year introduced ChatGPT Go, its under-$5 plan aimed at attracting Indian users, and later made it available free for a year in the country.

Similar dynamics are playing out for Anthropic: its Claude app recorded a 48% increase from the previous year in downloads in India in September, reaching about 767,000 installs, while consumer spending surged 572% to $195,000 for the month, per Appfigures — still modest compared with the U.S., where September spending hit $2.5 million.

Anthropic has been stepping up its engagement in India at the highest levels. Chief executive Dario Amodei visited in October and met corporate executives and lawmakers, including Prime Minister Narendra Modi, to discuss the company’s expansion plans and growing adoption of its tools. Anthropic had also explored a potential partnership with billionaire Mukesh Ambani’s Reliance Industries to broaden access to Claude, as TechCrunch reported previously. Reliance, however, ultimately struck a deal with Google to offer its Gemini AI Pro plan free to Jio subscribers. That move came as rival Bharti Airtel partnered with Perplexity to bundle access to its premium subscription, underscoring how India’s telecom giants have become critical distribution gatekeepers in the race to scale consumer AI services.

In a LinkedIn post announcing the move, Ghose said she would focus on working with Indian enterprises, developers and startups adopting Claude for “mission-critical” use cases, pointing to growing demand for what she described as “high-trust, enterprise-grade AI.” She added that AI tailored to local languages could be a “force multiplier” across sectors including education and healthcare — signaling Anthropic’s intent to deepen adoption beyond early tech users into larger institutions and the public sector.

The push by Anthropic, OpenAI, and Perplexity comes as India’s homegrown GenAI ecosystem remains relatively early-stage. While the country has a deep pool of software talent and a fast-growing base of AI users, it has produced few startups building large foundation models, with investors instead largely backing application-layer companies rather than committing the scale of capital typically required to train frontier systems.

The appointment also comes ahead of India’s AI Impact Summit 2026 in February, where the Indian government is expected to bring together AI startups, global CEOs, and industry experts to discuss the next phase of AI deployment in the country. The summit is part of New Delhi’s broader effort to signal support for domestic AI development and position India as a serious player in the global AI landscape, as competition intensifies across major markets.

Anthropic is also building out its India team, with job listings for roles including startup and enterprise account executives as well as a partner sales manager, signaling a push to deepen its go-to-market efforts and tap Indian businesses and startups as customers as it expands its presence in the country.

For Anthropic, the hire adds senior local leadership as it looks to turn India’s surging usage into a durable business, navigating a market where distribution partnerships, pricing pressure, and enterprise adoption will shape which AI players emerge as long-term winners.