Broadcom: The $600 Billion AI Chip Giant

Broadcom: The $600 Billion AI Chip Giant

Show Video

Broadcom is the second largest  AI chip company in the world. Thanks to that, the company is the  11th largest company in the world.   Over $600 billion as of this writing,  bigger than Visa and just behind TSMC.

It is a bit crazy considering that  in 2009 the whole company was worth   $4 billion. 150 times growth  in 15 years is kind of wild. But what is Broadcom? How did they get  this big? What do they do? In this video,   how a little chip division grew to  be a $600 billion AI juggernaut. ## Beginnings The company now known as Broadcom  started as a spinoff of a spinoff. In March 1999, the iconic California-based   computer-maker Hewlett-Packard  decided to split into two. Everything unrelated to computers,   IT or printers would be put into the  new company - Agilent Technologies.

The new publicly-traded company covered HP's  former test and measurement, medical products,   chemical analysis, and semiconductor businesses.   Accounting for about $8 billion of  HP's $47 billion total revenues. Analysts saw this as a necessary  re-focusing of a business that   had grown too large. In an interview at the time,   Agilent's new CEO Ned Barnholt positioned it  as a coming-out of sorts from the shadows. Agilent's debut on the markets was one of the   biggest IPOs in history up until  then. The stock popped nearly 70%.

But the new company struggled to grow in the rough  years after the bursting of the Dotcom and fiber   optic bubbles. The company's revenues  shrank nearly 50% from 2000 to 2001. And soon they started laying people  off and selling entire divisions to   raise money and simplify the  business. So in June 2005,   they put up their chip division - called the  Semiconductor Products Group - up for sale.

## Spinning Off the Chips There followed a brief auction... Which PE firms KKR and Silver Lake Partners  won for $2.65 billion in August 2005. The government of Singapore also co-invested  in the deal through their Temasek and GIC   sovereign wealth funds. Singapore had been  a partner with Hewlett-Packard since 1970,   when HP first chose to set up a factory there. It was a glorious time to be in private equity.  Some of the biggest private equity deals in   history were closed during this 2006-2007  period before the Global Financial Crisis.

Big deals included the utility company TXU (bad  deal), the financial services company First Data,   and the hospital company HCA. These  regularly topped $20 billion in total value. This Agilent carve-out deal was nowhere as  large as those monsters. That same month,   KKR closed a $10 billion deal for a  majority stake in Philips Semiconductor. That was later merged with Motorola's internal   semiconductor division Freescale  to become what is now NXP. Nevertheless, the Agilent deal  seemed like a good arrangement.   KKR and Silver Lake believed that the  division had the potential to thrive.

Hewlett-Packard's semiconductor  division dates back to 1961 when   the electronics giant believed that it had  to bring component production in-house. Over the years, that group expanded to a  range of industries. Some of their more   prominent products were FBAR RF filters used  for mobile phones’ radios (more on that later); Optical transceivers; And custom chips for HP's lucrative printers  to lock you into their ink cartridges; But the semiconductor business changed  a great deal in the 1990s. The rise of   foundries turned the tables on the economics of  owning your own internal semiconductor division.

And despite an admirable attempt to  diversify, the division’s cyclical   nature and chronically low margins exposed  Agilent to losses when times were bad. 17%   of the semiconductor division's 2004  revenues still depended on HP itself. In the end, Agilent's management  realized that this business was   not "core". Thus in December 2005, Avago  Technologies emerged into the world. It  

was the largest privately-held fabless  semiconductor company in the world. ## Hock Tan Avago's first CEO was the division's  original general manager Dick Chang. Chang originally joined Hewlett-Packard Labs and   became Vice President and General Manager  of the Semiconductor Products Division. But a few months after its founding,   Avago announced that it had a new president  and chief executive - Hock Eng Tan (陳福陽). Hock Tan is of Chinese descent,  born in Malaysia. At the age of 18,   he won a scholarship to MIT, where he  received a bachelor's degree. Then he  

went to Harvard Business School to do an  MBA. After that, he returned to Malaysia. There, he worked at a number of companies. First,  Hume Industries, a cement maker affiliated with a   large Malaysian conglomerate. Then a venture  capital firm called Pacven Investment.

After that, he returned to the United States.  There, he took finance roles at Pepsi and   General Motors before landing at the personal  computer company Commodore International in 1992. At Commodore, Tan served as vice president of  finance and then CFO during the iconic computer   company's final days as it struggled with huge  losses and low sales. That must have been brutal. After Commodore announced its bankruptcy in 1994,   Hock Tan left to become SVP of Finance at a chip  company called Integrated Circuit Systems or ICS. ## ICS ICS was founded back in the late  1970s in Valley Forge, Pennsylvania. The company was a design house, meaning that  they designed semiconductor products but did   not sell them. They did contract design work for  larger companies like United Technologies or GE.

In the late 1980s, they developed a  hit new product - a silicon timing   device or frequency timing generator. Every  electronic system needs a timing element,   and many older ICs used to  use crystals to do that. The silicon timer replaced those  crystals with an all-silicon solution,   and that was a big deal. It was a huge  success and in 1991 they went public.

Hock Tan joined ICS in 1994. Then in 1995, he  became CFO. A year after that, he became COO too. Then in 1999, ICS announced a leveraged buyout  by its own management team for $257 million. The   PE firm Bain Capital and the bank Bear Stearns  helped provide funding. Hock Tan became the CEO.

In 2005, ICS merged with another company  producing mixed signal chips - Integrated   Device Technology in a $1.7 billion deal.  Tan became the combined company's chairman.   It was then that he was recruited  by Silver Lake to be Avago's CEO. ## The Franchise Bloomberg quoted Kenneth Hao; A partner at Silver Lake and now chairman  and managing partner - saying about Tan: > He starts with a point of view  that the semiconductor industry   has matured ... The businesses must be run  differently than when they were growing up Tan really likes the idea of focusing  on a company's core "franchises" - a   word he has used multiple times. In a 1999  interview while he was still CEO of ICS,   the silicon timer company, he  likened ICS to a franchise: > We have essentially in our business almost  a franchise ... in the sense that anybody who   wants timing solutions ... would immediately  think not just of crystals but of silicon,  

and not just silicon but of  Integrated Circuit Systems > Our market share in the PC space is  extremely high, and that effectively   gives us almost a franchise ... We get to be  selected in most situations as the first source Years later, he brings up the phrase again in  a 2018 interview with the Wall Street Journal: > It's about putting together a very good  portfolio of product franchises to create a   lot of value ... There’s no long-term vision  or ambition other than that ... Frankly,   we overinvest to ensure we are  way ahead of No. 2 or No. 3 The franchise metaphor is illustrative. The  best competition is no competition. Fast   food and car dealership franchises maintain  small localized monopolies over their areas.

In the old days, the industry's  cyclical nature and early growth   motivated the old guard semiconductor  companies like Motorola and National   Semiconductor to launch as many products  in as many different fields as possible. But things have since gotten  more complicated. Designers are   now often more interested in solutions  for their systems rather than bespoke   components. And foreign competition  has flooded the market with options. So Tan's thinking is that today's leading  semiconductor businesses need to build   and maintain "franchises". To be the  instinctive number one choice whenever   a designer needs a thing for their  system. A small localized monopoly. ## Strengthening the Franchise When Tan takes on a new business, he wants to  strengthen that "franchise" as much as possible.

That means a focus on cutting-edge products  within the franchise as well as new verticals   and applications to expand into. When  asked about ICS's growth strategy he said: > We’ll stick to our knitting, essentially,  which is silicon timing solutions. Our strategy   is to expand applications, and increase market  share in digital consumer and communications This means cutting speculative projects  like the ones the old guard semiconductor   companies used to do. If it does not  help the core franchise, chuck it. This approach has more than a few critics. Cuts to   R&D expenditure or sales of non-leading  divisions have long term implications,   since these things take time to bear fruit. The  franchises have to come from somewhere, right? Tan correctly points out that he invests  considerable resources into existing   franchises - milking the cow so to speak.  Nevertheless, a company of Avago's style  

must - like a hummingbird - be eating  constantly in order to appreciably grow. ## Road to IPO Avago began life as a very diverse business. At the end of 2005, the company had five  different end markets generating over 10%   of sales. This was the convoluted legacy  of forty years of diversification at HP. So Tan and his team started organizing  this mishmash to create a coherent   company focused on its analog, mixed  signal, and optoelectronics franchises.

In March 2006, they sold their storage  business to PMC-Sierra for about $420 million. In May 2006, they sold off their printer ASIC  business to Marvell for about $250 million. Then in December that same year  they sold their CMOS image sensor   business to Micron Technology for $53 million.

They did one more sale a year later  - selling their Infrared operations   to the Taiwanese electronics  company Lite-On for $20 million. These sales not only raised  money to pay down debt,   but also slimmed the company's headcount  from 6,500 in 2005 to around 3,600 in 2008. By 2008, though Avago's revenue and earnings had  not changed all that much since the acquisition,   the company had paid off a  billion dollars of net debt. The improvement was enough for  KKR and Silver Lake to take the   company public in August 2008, two  years after it first went private. The two PE groups would later sell down  their shares in the years after the IPO,   eventually realizing a 5x return on their  shares. Not bad for a couple years' work. ## The Mobile Revolution The IPO came about at a fortunate time.

2007 and 2008 heralded the introduction of  the Apple iPhone and the subsequent smartphone   mega-boom. And Avago parlayed their early work in  FBAR RF filters to make the most out of the surge. As I detailed in a prior video, RF  filters are critical but unheralded   parts of the mobile phone. They  help the modem separate the data   signal from noise - saving on power  and improving the user experience.

These filters are actually more like MEMS  - tiny tuning forks - rather than Nvidia   GPUs. But they are highly specialized -  requiring special thin film techniques   for certain unconventional  materials like gallium arsenide. As I mentioned earlier, HP had been  working on FBARs since the 1990s. In 1999,  

they released their first commercial FBAR filter. Two years later in 2001,   they released the first FBAR filter for  the 1900 megahertz band, a GSM band. In 2008, Avago paid $30 million USD  to acquire Infineon's BAW RF acoustic   filter business and their patents.  FBARs are one major subtype of BAWs   filter. Infineon's filters  were SMRs, another sub-type. In 2010, Avago was first to market with  an RF filter for the 4G-LTE bands. LTE  

sparked a mobile data revolution across  the world, forcing RF filters to have   to accommodate increasingly more  bands PLUS Wi-Fi and Bluetooth. This greatly increased their complexity,   and thus their prices. The RF filter industry  grew from about $100 million in 2004 to over a   billion in total revenues a decade later.  Avago earned a good chunk of that money. At the same time, smartphones got  skinnier and more power efficient.   This necessitated the consolidation of  the many various discrete components   of what we call the RF front end -  power amplifier, filter, switches,   and antennas. Everything short of the transceiver  and modem, which is dominated by Qualcomm.

So the mobile smartphone and LTE booms  were good for Avago throughout the late   2000s and early 2010s. But by 2013  mobile was some 50% of their revenue,   raising concerns of over-concentration. So Hock  Tan and Avago went out to buy a new franchise. ## LSI & The Data Center In December 2013, Avago announced that it  would acquire LSI Logic for $6.6 billion. I covered LSI Logic and its journey in a  prior video. What follows is a brief summary. LSI had been founded by the former  CEO of Fairchild Semiconductor - a   no-nonsense Brit named Wilf Corrigan.

They raised a little money, engaged a  Japanese fab to make wafers for them,   and came out with an innovative “master  slice" system that helped customers   replace entire circuit boards with custom chips. Revenues exploded to nearly $2 billion.  However this space had many competitors   and times got tough after the Dotcom bubble burst. After Corrigan retired, LSI's new CEO shifted  to a fully fabless business model and started   developing custom silicon for running data  centers and helping to make them more efficient. LSI Logic already had a foothold in this  business through their work on hard drive   controller chips. They then bulked it up in 2007  with a $4 billion merger with Agere Systems. Agere was a spin-off company from Bell Labs/Lucent   that made semiconductors and components for  wired and wireless communications networks.

Together with Agere's business,  LSI Logic built up a proficiency in   enterprise storage systems - networking and  storage hardware connected to SSDs or HDDs. In a time when cloud storage startups  like Box and DropBox were raising   at billion dollar valuations, this  franchise attracted Hock Tan's team. And for LSI, they had been in the  desert for a few years anyway,   a payout would be a good end for them. Avago's acquisition of LSI Logic was like one of   those deep sea gulper eels that try  to swallow fish way bigger than them.

In 2013, Avago produced $2.5 billion in revenue.  Their market cap was just about $12 billion. Only $1 billion of the $6.6 billion  purchase price was their own cash.   Their PE partner Silver Lake invested  $1 billion and the rest, $4.6 billion,   was borrowed. So again, Hock Tan and his team  started cutting expenses and paying off debt.

In May 2014, they sold LSI's flash memory  solutions - which includes their PCI Express   flash solutions as well as controller chips for  solid state drives - to Seagate for $450 million. Then in November 2014, they  sold LSI's network chip unit,   called Axxia, to Intel for $650 million.  The unit's chips help internet service   providers and data centers monitor  and manage traffic over networks. ## Broadcom & Brocade A year after that, Avago struck another, far larger deal - the venerable  analog chip company Broadcom. Avago had tried once in late  2014, but couldn't agree on price. But the LSI acquisition drove up Avago's stock  and gave it a richer currency to do a better deal.

So in April 2016, Hock Tan called up Broadcom's  management again with an offer, then tried to   hash out a deal right then and there. From  start to end, it took a month and a half. The $37 billion acquisition was the biggest  technology deal up until then. Avago paid a   nice 28% premium on Broadcom's share price,  which was then already at a 9-year high.

Broadcom is a storied name in semiconductor  history. Founded in 1991, they produced analog   system-on-chips for high speed communications  equipment like cable modems and TV set-top boxes. So throughout the 1990s, they grew  with the rise of cable operators   like Comcast and broadband internet. In the 2000s,   they entered a series of other markets  like enterprise switches and networking. Broadcom was another gulper  eel type deal. Its revenues   in the prior year were nearly  twice the size that of Avago's.

As part of the deal, Avago agreed to take  Broadcom's name and add Broadcom co-founder   Henry Samueli to its board. It reminds me  of those type marriages between a new-money   family and a family that is poor but with  a long and prestigious name plus a castle. Little over a year later, Broadcom acquired  another storage networking company Brocade   for about $6 billion. Again the cycle restarted,   cutting expenses and paying off  debt. We don't need to rehash this. ## Qualcomm So in just a few years, Avago/Broadcom  gathered a great bunch of assets. It is a big company version of buying properties  on mortgage, fixing them up, renting them out,   and then paying down the debt aggressively.  Rinse and repeat for a few times over.

The issue though was that after Broadcom  and Brocade, there were not many targets   left big enough to move the needle. But  you miss 100% of the shots you don't take,   so Tan went right for the prom queen - Qualcomm. I also covered this saga in a prior  video, but this is the summary. On November 6th 2017, Broadcom offered $103  billion - an offer immediately rejected by   the Qualcomm board as too low. They tried  to engage but was rebuffed - so they went   direct to Qualcomm shareholders in  an attempt to swap out the board. In January 2018 Qualcomm filed a notice  with CFIUS to block the acquisition.  

CFIUS is a committee overseeing  national security issues on mergers. CFIUS highlighted concerns about  Broadcom's reputation of cutting R&D,   feeling that it would extinguish  Qualcomm's chances of competing in   5G technologies against Huawei  and other Chinese companies. In response, Broadcom re-domiciled itself  out of Singapore to the United States.  

Then-President Trump held a public  conference with Hock Tan on this. The re-domiciling would have presumably  invalidated CFIUS's jurisdiction. But the Trump   Administration ended the whole kerfuffle when  it issued an executive order blocking the deal. That happened six years ago - time really  flies. Qualcomm's market cap has since risen  

to a number above Broadcom's offer price due to  surging revenues from the COVID technology boom. But the stock price hasn't moved all that much  since in part due to the post-COVID hangover,   Huawei competition, and what not. As of this  writing, Qualcomm's market cap is $190 billion. Looking back, I feel like the  Qualcomm shareholders would have   preferred the growth from having Broadcom shares.

But on the other hand, CFIUS's  point was that they wanted to   retain Qualcomm's ability to compete  in certain wireless markets like 5G.   Broadcom would have supposedly snuffed  that out. It wasn’t a money thing. Yet Huawei is still dominant in 5G and Qualcomm's   modem work seems to be more in the news  for fighting Apple on royalty payments. But I certainly feel that the Broadcom  acquisition would have snuffed out   Qualcomm's attempts to compete in Arm-flavored  laptop chips like Snapdragon X. On the whole,  

I still think it's a good thing  that Qualcomm stayed independent. ## The AI Boom Blocked from further growth in  silicon, Hock Tan and Broadcom   have since turned to acquiring software companies. This includes CA Technologies, Symantec, and  VMWare. Basically running the same processes,   pulling out the franchises, cutting expenses, and  paying off whatever debt incurred for the buy. But until ChatGPT came out, there were always  questions of what was next. What thing can they   get next to take them to the Holy Land? The  answer had been inside them all this time.

When Avago/Broadcom acquired LSI Logic in  2013, they also acquired a small custom silicon   design division that helped external customers  produce their own chips for the data center. When Google started on the first  iteration of their own chips - the   Tensor Processing Unit or TPU - they did  not need that many of them. So in 2016   they engaged Avago/Broadcom to help  design and produce the chip for them. Why did Google do this? Because custom silicon  design teams are expensive. You not only pay a   lot for talents, but also EDA tools and IP for  them as well. Unless your volumes are massive,  

it is better to engage LSI or Avago  to do the design based on your spec   plus other value-adds like testing and packaging. Back in 2016 when the TPU relationship presumably  began, custom silicon was quite small for both   sides. Broadcom's revenues from that year  was estimated at about $50 million or so. But it grew over time and by 2020, Broadcom’s TPU  revenues were estimated at $750 million. Together,   the two companies have made several  new TPU iterations on faster nodes. When ChatGPT blew up in November  2022, it started the current AI boom,   with everyone from startups to tech giants  getting into generative AI. Many of these   players hit compute bottlenecks, because nobody  not even OpenAI expected ChatGPT to be such a hit.

Google realized that their TPU ASICs gave them  a huge competitive advantage in compute over   Microsoft, Oracle and others. So they started  buying more TPUs, which benefits Broadcom. SemiAnalysis projects that Google  will pay Broadcom far more in 2024   for TPUs - something like $8.5 billion. Only some   lame computational lithography company  called Nvidia makes more from AI chips. That is not counting whatever they make from their   relationship with Meta or  other giants like Microsoft. But Nvidia isn't going to let Broadcom have  that all to themselves. In February 2024,  

Reuters reported that Nvidia is setting  up their own AI chip design unit. The   idea being to leverage their IP  and expertise to keep customers   from designing alternatives to their  profitable H200 and B100 accelerators. Good. Let them fight. ## Conclusion I want to thank Dylan Patel of  the aforementioned SemiAnalysis   for some clarifying comments on this video.

Dylan is great. He did a piece prior  on Broadcom's history that I think   would be good companion reading to this one. So what is Broadcom? Friend of the show  Digits to Dollars lays it out pretty clear:   Broadcom is a publicly traded private equity  fund masquerading as a semiconductor company. They have a portfolio of strong technology  franchises. They acquire new ones,  

slim them down, focus them, and then leverage them   to buy new franchises. They grow  whenever they buy something new. This time, one of their franchises has stumbled  into the heart of the AI boom. That does not mean   they won't fight like heck to stay there like  with the mobile and cloud booms prior. I reckon  

they will invest whatever necessary to keep  and extend their position in this gold rush.

2024-04-06 20:31

Show Video

Other news