After a bright financial report, the stock price plummeted? Broadcom Inc. (AVGO.US) CEO: Never caring about stock price, sticking to technology customers will not run, bullish on the "Broadcom Inc.-Alphabet Inc. Class C-Anthropic" triangle.

Despite the record-breaking second-quarter financial report failing to meet Wall Street's "extremely high expectations" for AI and causing a significant drop in stock prices, Broadcom Inc. CEO Hock Tan responded with extreme confidence in a recent interview: "We never care about stock prices," and for the first time revealed the timeline for production in collaboration with OpenAI on AI chips, as well as the astonishing returns on the "Broadcom Inc.-Alphabet Inc. Class C-Anthropic" ecosystem. At the 2026 Bloomberg Technology Conference held in San Francisco on June 5th, Broadcom Inc.'s President and CEO Hock Tan sat down for an exclusive interview with Bloomberg reporter Tom Giles to discuss semiconductor demand, the expansion of AI scale, and revenue prospects. This came just after Broadcom Inc. announced strong second-quarter results for the 2026 fiscal year, but the stock price saw significant selling. According to a previous article from Wall Street News, in the latest financial report, Broadcom Inc.'s second-quarter net revenue exceeded $22 billion for the first time in a single quarter, representing an approximately 48% increase year-on-year; AI semiconductor business revenue saw an almost 80% year-on-year surge. However, due to the company's third-quarter AI semiconductor revenue guidance ($16 billion) being slightly lower than some analysts' "extremely high expectations" of over $17 billion, this AI giant stock, which has skyrocketed in market value by over $300 billion since the beginning of the year, saw profit-taking after hours, with the stock price falling by over 10%. In the midst of the market's near-perfect expectations for financial reports in the "AI frenzy period," Hock Tan displayed the calm demeanor of a seasoned tech giant leader in the interview. Hock Tan bluntly stated, "We are in a very surreal AI environment. Honestly, I never think about it (the market's extremely high expectations). It's hard not to pay attention now, but I only focus on the fundamentals, focus on creating immense value, and no longer think about stock prices. Focusing on stock prices only brings trouble, although it's hard to do, we are trying to do it (not look at stock prices)." "Doubling down on technology," constructing the "Broadcom Inc.-Alphabet Inc. Class C-Anthropic" triangle In the field of AI custom chips (ASICs), Broadcom Inc.'s deep integration with Alphabet Inc. Class C on TPU is its most critical cornerstone. Faced with concerns that giants such as Alphabet Inc. Class C seeking "customer-owned tools" (i.e., fully independently developed chips) may threaten Broadcom Inc.'s position, Hock Tan appeared unfazed. "We've been in the semiconductor industry for over 20 years... Our goal is to exceed any competitors through engineering technology." Hock Tan pointed out that Alphabet Inc. Class C is trying to achieve self-design with the help of smaller-scale partners, but he believes the core logic remains in technological competitiveness. He pointed out that the irreplaceable reason for Broadcom Inc. lies in the steady pressure from NVIDIA Corporation: "The real competitor Alphabet Inc. Class C faces is NVIDIA Corporation's GPU. As long as NVIDIA Corporation continues to release generation after generation of superior technology, Alphabet Inc. Class C must create equivalent technology to compete. This is where we (Broadcom Inc.) shine." In addition to the underlying hardware, Broadcom Inc. has completed a near-perfect precise investment in the AI large model ecosystem. About a year ago, Broadcom Inc. collaborated with Alphabet Inc. Class C to provide computing power support for the emerging AI giant Anthropic through a jointly designed TPU. Hock Tan described this as a "leap of faith." "This is a great bet," Hock Tan said, emphasizing that Broadcom Inc. not only sees potential in Anthropic but also in the explosive growth of generative AI in the enterprise sector. He revealed that Broadcom Inc. is internally using Anthropic's Opus 4.7 and other large model tools for engineering design and code assistance, with an exceptionally remarkable return on investment. "When engineers become proficient in using these tools, you can have an excellent senior engineer complete an application design in a week; in the past, this would have taken ten engineers with an annual salary of $300,000 each three months to complete. The return on investment is very substantial." Dispel rumors: OpenAI custom chip "to go into production by the end of the year" Last year, the market highly anticipated Broadcom Inc.'s historic collaboration with OpenAI on AI chips. Addressing recent rumors that the collaboration may have encountered obstacles or required Microsoft Corporation's approval to purchase a certain percentage of chips, Hock Tan categorically denied these claims in the interview. Hock Tan revealed for the first time the client map in the field of custom AI accelerators: "Specifically, we have only 6 (custom chips) clients, each at different stages of creating their own silicon chips. Obviously, Alphabet Inc. Class C is at the forefront. OpenAI is also one of our collaborating clients, with whom we have been working closely for over two years." Regarding the progress of the OpenAI chip, he provided a clear timeline: "We have achieved remarkable results. The AI accelerator chip runs very well in their labs and data centers. We are on track and will go into production later this year. Everything is ready, and we are about to take off." When asked by the host if any agreement with Microsoft Corporation was needed to advance the collaboration, Hock Tan firmly replied, "Not at all." "Selling water" to earn a full bowl, no plans for mergers and "chasing shiny objects" Historically, Broadcom Inc. has been known on Wall Street for its aggressive and precise mergers and acquisitions (M&A). However, faced with the surge in AI compute power demand sparked by generative AI, Hock Tan made it clear that the company has entered a new phase where it no longer needs to be driven by M&A. "This is a very realistic question," Hock Tan did some math, "Between 2024 and 2026, my revenue will double, creating over $50 billion in revenue annually. When I look around, what can I acquire that can compare to that? There's nothing." He emphasized that the insatiable demand for compute power in generative AI makes organic growth already astonishing enough. At this juncture, engaging in mergers and acquisitions, dealing with regulations, and integration is a distraction. "I have been operating under this business model for 20 years, and I am working hard to avoid chasing those 'shiny objects' (referring to blindly acquiring popular concepts)." Discussing Broadcom Inc.'s dominant position in network switching equipment and facing competition from companies like Cisco Systems, Inc., Hock Tan maintained extreme confidence. He stated that Broadcom Inc. is not only the market leader in about 5 to 6 core sectors, including networking, but also a key player in building AI clusters. "The current demand for AI and AI networks is enormous. We will prioritize providing products to those strategic customers that have sustainable development plans," Hock Tan supplemented with a smile, "As for those orders overflowing (our capacity), I am happy to pass them on to Cisco Systems, Inc." The full transcript of the Bloomberg interview with Broadcom Inc. CEO is as follows: (Bloomberg Tom Giles: Okay, so what does this indicate? Investors have such high expectations for companies like yours - you released a financial report, reaffirmed forecasts for revenues reaching billions, surpassing market expectations in many ways, but there is still a sense of disappointment in the market. What stage do you think we are at in the AI hype cycle? Broadcom Inc. CEO: I think we are in a very surreal AI environment right now. Honestly, I usually don't think about these things. So it's hard not to think now, but now you just need to focus on the fundamentals, focus on creating real value, not always think about your stock price. Stock price issues are indeed hard to ignore, but we are trying to do it as much as possible. Bloomberg Tom Giles: Like we just talked about with Ali, you know, about every month or every few months, there is a new large language model coming out, which in a way changes the whole industry discussion about who is leading and who is falling behind. I'm curious, you have a large-scale partnership with Alphabet Inc. Class C, a very close collaboration on TPU as you mentioned. I want to dive deeper into this topic with you. Specifically, how concerned are you about Alphabet Inc. Class C bringing more chip development design work in-house? The concept of "customer-owned toolchain," will it keep you up at night? Have you set any clauses in the collaboration agreement to guard against such occurrences? Broadcom Inc. CEO: This area is one that we need to continuously expand in from a technological perspective. We have been in the semiconductor industry for more than 20 years... In essence, this is what I have been doing during my tenure at Broadcom Inc. We have multiple product departments, and in each area, we are the best, we continue to invest in research and development. Essentially, what you need to do is to surpass any competitors through engineering technology. Alphabet Inc. Class C is trying to advance the "customer-owned toolchain," meaning they want to design their own chips and rely on some smaller-scale partners to achieve this goal. Bloomberg Tom Giles: Such as Marvell? Broadcom Inc. CEO: Companies like Marvell. I call them "guidance type" partners, they basically serve Alphabet Inc. Class C, helping Alphabet Inc. Class C achieve customer-owned toolchains. So we are competing with our own customers on another level. The crux of the whole logic is that we have to differentiate our products and technologies, surpass what they are creating themselves. This is advantageous to us, largely because the real competitors these solutions face are NVIDIA Corporation's GPUs. As long as NVIDIA Corporation keeps releasing one generation after another of outstanding technology, Alphabet Inc. Class C must create equivalent technology to compete, and this is where we can make a difference. Bloomberg Tom Giles: Right. The development process of TPU is quite complex, I'm curious if there are any ways, or what you can do, to help Alphabet Inc. Class C make this process more accessible? Broadcom Inc. CEO: What do you mean by "accessible"? Bloomberg Tom Giles: I mean, making TPU available to more mainstream developers, not just limited to top AI labs. Broadcom Inc. CEO: Well, I think Alphabet Inc. Class C is opening up the TPU environment through cloud services - Alphabet Inc. Class C Cloud GCP - to achieve this goal. But an important part of this is the work at the software level, including operating systems, software, and accompanying compilers, which are core businesses of Alphabet Inc. Class C. We are certainly happy to help, but I believe they are quite mature in advancing this project and have essentially controlled the situation. Bloomberg Tom Giles: I'd like to talk about Anthropic. It is clear that the builders of these foundational models need a lot of capital to support their chip and compute needs for developing and iterating their models. Could you discuss how this partnership was built and financed? What safeguards have you taken to mitigate potential risks? How concerned are you about the ability of end users to truly land their commitments and fulfill them? Broadcom Inc. CEO: Frankly, in this specific case, we are a partner with Alphabet Inc. Class C, providing computing support for Anthropic through the TPU we co-designed with Alphabet Inc. Class C. To be honest, there is no shortcut to this. About a year ago when we just started the collaboration, it was basically a leap of faith, a bet on Anthropic and generative AI. We believed that Anthropic, through its business model focusing on serving enterprise clients, providing code assistance and programming tools, could bring about real change. We took this bet, and based on what we know now, it was a very correct bet. There are valid reasons for this, though a year ago, what I knew may have been just a fraction of what I know now. And what I know now is probably just a fraction of what I will know in six months. This all changes very quickly. But from what we see at Anthropic, and also from tools coming from OpenAI, as well as Alphabet Inc. Class C Gemini, their tools can dramatically increase productivity. We ourselves are using these tools for engineering design, code assistance, and the results we can achieve in enhancing productivity - either in improving my personal productivity or product development productivity. I even believe that the design solutions created by these tools may be better than what our engineers could come up with independently. And our engineers are indeed very talented. Bloomberg Tom Giles: Can you elaborate on how you guide employees to use these different tools? Have you found a need to limit this? We have heard of a trend of "tokens being stretched to the limit" - I don't particularly like that term - meaning allowing unrestricted use of tokens. Have you set usage limits? How specifically have you done it? Broadcom Inc. CEO: No, we haven't set any limits. Frankly, from a more macro perspective, we have not reached that stage yet. Perhaps it is because we started later than others, and have not reached that point yet, so we have not begun to set limits on usage. An important point here is: when engineers use tools like Opus 4.7, they go through a learning process and learn how to use it well. Initially, you cannot use the tools very efficiently, but as long as you persist in using them, you will become more proficient, and the efficiency of the tools will increase significantly. As your efficiency increases, consider this: you can have an excellent senior engineer complete an application design in a week. In the past, the same task might require ten engineers, each with a $300,000 annual salary, to spend three months to complete. The return on investment is still quite substantial. So, whether to "throttle" is actually a question of return on investment and depends on what application you are working on. At the same time, it is an ongoing learning process and getting better at using the tools. Bloomberg Tom Giles: Let's talk about the OpenAI collaboration. Have there been any setbacks in this collaboration now? Whether it needs Microsoft Corporation to agree to purchase a certain number of chips to complete the agreement, how is this progressing? Broadcom Inc. CEO: Oh, it's not like that. By the way, about developing chips in collaboration with partners - take the example you mentioned earlier, my collaboration with Alphabet Inc. Class C has been very cooperative. They are trying to customize tools, and the whole process is very friendly and harmonious, everyone aims to do things well. We have a deep partnership with them, just like we do with most major clients. But for them, it's a journey. The first generation of chips will be replaced by the second generation, and it will continue to improve. So for them, it's a journey - developing custom AI accelerator chips, optimizing them fundamentally for specific workloads, ultimately replacing the common solutions currently in use. We are doing this, and we are also focused on this. At present, we only have a few clients - six of them - each at different stages of creating their own custom chips. Obviously, Alphabet Inc. Class C is ahead of the pack, leading the way. OpenAI is one of them, and we have been in collaboration for over two years. The results we have achieved are quite remarkable - our AI accelerator chips run very well in their labs and their data centers. We expect to officially go into production later this year. In this way, we are on full throttle. Bloomberg Tom Giles: So, is there currently no obstruction in this collaboration? Whether it's needing Microsoft Corporation to reach a certain agreement or for any other reason? Broadcom Inc. CEO: Not at all. Bloomberg Tom Giles: Okay. Let's talk about mergers because historically Broadcom Inc. has considered mergers as a significant growth strategy. With the arrival of the AI revolution, will you reduce the need for growth through mergers? Are you entering a "post-merger era"? Or is this understanding wrong? Broadcom Inc. CEO: No, Tom, you hit the nail on the head. This is a very interesting question, and a very valuable question. This question often pops up in my mind, and is a topic that has been continuously discussed at my board meetings - whether we should continue to do mergers and acquisitions. Let me give you a straightforward answer. In the past two years, between 2024 and 2026, with my revenue doubling, creating over $50 billion in revenue annually. When I look around, I ask myself, what can I acquire that can compare to that? There's nothing. And mergers themselves are a distraction - mergers themselves are a distraction, and then there's going through the regulatory process, which is another distraction, then spending a year integrating, yet another distraction. At the same time, with organic growth, the phenomenon of generative AI we are talking about, and our ability to deliver "picks and shovels" (i.e., computing power) to meet almost insatiable demand, it makes it very difficult to turn to mergers and acquisitions, rather than focus on succeeding in the field of generative AI computing. Bloomberg Tom Giles: Is there a field where you think you might make an exception - like photonics, just a random example? Broadcom Inc. CEO: Photonics? Do you mean optics? Bloomberg Tom Giles: Just throwing it out there. Broadcom Inc. CEO: I have been running this company for 20 years. I have been very persistent in... you know, avoiding chasing those "shiny objects" - as I mentioned before. Bloomberg Tom Giles: Got it. I want to ask you some similar questions I asked other keynote speakers. You know, when you see the price sensitivity around token usage, and the price sensitivity in decision-making by enterprises, considering such expenditures. We have seen many companies starting to shrink, limit the use of tokens, and so on. As we discussed before, we have also seen signs that some large language models in China are beginning to receive a lot of attention, especially considering how efficient they run, whether it's DeepSeek or Alibaba Group Holding Limited Sponsored ADR's Qwen. From your perspective, have you observed this situation? Broadcom Inc. CEO: I, from my perspective, this is what I have heard. I can't say if we are doing this internally at Broadcom Inc., but I have heard a lot - I think you have heard from the same sources - that many corporate leaders, CIOs are talking about, "Hey, you know, we are trying to reduce the cost of tokens," and things like that. Maybe they are ahead of us in this respect. However, my judgment on the use of AI by enterprises is: we are still in the early stages of this game. I believe we are still trying to figure out what we can achieve from these magical tools we have created and continuously enhanced and improved with each successive generation. So from my personal viewpoint, as I mentioned before, and from my experience managing Broadcom Inc., yes, I don't want my team to spend money carelessly on tokens. But on the other hand, if you can create value from it, and you can generate an investment return that was previously unattainable, then why stop? It's not like going to an arcade - as we played games when we were kids, when the monster was about to eat you, and your gun ran out of bullets, so you had to put in more coins. The situation is almost like that, but not entirely.