Venture Capital funding in semiconductors going up…

I was reading through the 2006 Venture Capital Industry Report and found a very interestin, and very heartening, news.

(Pg. 10)

The number of companies and amount of capital
invested in semiconductors and electronics has grown
significantly, reflecting increasing activity in storage,
mobile, and consumer devices.

The median amount invested in semiconductors is standing at bubble-era funding – $10m (Pg. 16)

(FYI – investment in healthcare and bio-pharma services are reaching MUCH higher levels than bubble-era funding)
It was also very interesting to note that the most active corporate investor was … Intel Capital

However, early stage companies were left to fend for themselves in the big, bad world

(Pg. 14)

As they have in recent years, investors also concentrated a
bit more heavily on established companies that were generating
revenues, rather than those in startup and development phases

Number of deals closed in Semiconductors is pathetically low (34 compared to 103 pre-bubble). However pre-money valuation has increased from 13.7 M to 20 M (I dunno if this is because the VC blokes have too much money to throw around, or has the general quality of the startups gone up??)

Now for the bad news – # of deals in Design Automation software has gone down to 27 from last year’s 39 (Pg. 25). In fact the only segment in the semiconductor software industry which is going up is Communications/Connectivity (???)

The EDA companies which successfully raised capital were also interesting – LVL7 Systems was the highest with $15M funding and Eridon the least with $1.7M funds (one more thing was that several of the EDA companies receiving funding were third/later rounds!!)

Compare with ASIC or General purpose IC companies – highest funding was $48M for SolarFlare and least was $4.65M for MultiSpectral Imaging.

Number of semiconductor companies going for IPOs in the US market(Pg. 56) was 4 (+ 14 M&A). Compare it with the European market where the IPO number was 47 !!!! Read that again – 47

The amount of funds raised by VC firms was 160% of 2004 fund-raising. Israel had 20 IPO’s in Semiconductors…

Where’s that book on STL…

del.icio.us Tags: vc business startup eda

Pitch and Presentation

Ashwin, over at Venturewoods, linked to a great presentation. This is a presentation given at OSCON 2005 by Dick Hardt, CEO, Sxip Identity.

I think the most important thing is to have a flawless demo (if any) and an intensity that burns through – everything else is just language.

I mean look at the Yoda Master himself.

del.icio.us Tags: personal business startup

27 % of what ??

A very nice article by Gabe Moretti pushes back against the caustic comments vilifying the EDA industry these days. It particularly refers to growing attacks, using the EDAC 2005 Market Statistics Service (MSS) report.

However, the most interesting part of the story was referring to an article in EETimes where the reporter quotes

Dataquest estimates that such in-house development investment increased to 27 percent, the highest since the inception of the commercial EDA industry in the early 1980s

That is interesting – more so because going by the DATE 2006 presentation by Dr. Wally Rhines (CEO, Mentor Graphics), it seems that equity funding in EDA companies has been increasing year on year (see page 10 and ignore the aberrant 2000 bubble-era funding).

Now, seems to me, 'tis very hard to explain the continuing naivete of venture capitalists – given that in-house development in increasing.

But I think it is very easy to explain – a large part of the charter for any in-house T & M (Tools & Methodologies group) is to define design flows and validate new tools. Defining a flow is how a tool actually gets used inside a design house. Now there are two reasons why EDA companies cannot do that:

• In EDA we do not trust - Design companies like to keep details of how they go about their design process a close IP. This includes workflows for engineers, testing, QA processes, etc. The last thing they would do is to bring in EDA companies into the loop, giving them tools to go and implement somewhere else. I personally think that this fear is unfounded – design flows have as much to do with the personality of the company as much as the technology behind it. Replicating one design process in another place is, in my opinion, impossible.
• Format Wars – OpenAccess, BLIF, Liberty…. and so on. All vendors have different formats – and not just to create a walled garden, but because their specific data structures are optimized for those formats.

27% of what is a good question – that has to be answered before the EDA industry gets panned.

del.icio.us Tags: eda business

Web 2.0 – Podcasting and Proaxys

This is a company started by one of my batchmates at IIT Bombay – Gaurang Kanvinde. Proaxys Software has technology to convert ordinary XML feeds to mp3 – which they embed in a feed of their own.

So basically, they do auto-podcasting or podcastify their feed.

Take a hear at three partial feeds that I podcastified from Robert Scoble's website (my apologies).

For example http://scobleizer.wordpress.com/2006/04/07/net-bcl-team-blogging/ is mapped to http://www.proaxsysreader.com/data/rss/2/1144521572771.mp3 . I liked the way their software recognizes ".NET" to be pronounced as "dot-Net"

Even more interesting is http://scobleizer.wordpress.com/2006/04/07/daily-kos-crashes-microsofts-gate/ mapped to http://www.proaxsysreader.com/data/rss/2/1144521574796.mp3. It can pronounce "Tamara Pesik" and "Markos Moulitsas Zuniga". That has got to count for something!!!

To try it out yourself -

1. go to http://www.proaxsysreader.com/rss.jsp

2. Enter the feed url and your email address

3. Check your email to get the auto-generated passord

4. Go to http://www.proaxsysreader.com/myreader/index.jsp

5. sign in

6. Click on "my feeds" and listen (using a podcast compatible software)

EDIT: Gaurang has mailed and said that the link generation process for your feed has been greatly simplified – sounds quite cool too!!
Technorati Tags: proaxys scoble podcasting

del.icio.us Tags: personal podcasting

EDA Wars: Return of DFM

I was going through one of John Cooley's slightly older posts, where he talks with two startups about DFM and what might happen in the world of EDA.

The interesting part that I saw was where everyone was speculating about who will have a consolidated DFM flow. Most of them hedged their bets with the Big Three – for no other reason, but the fact that they have an entire suite of tools.

Me? I bet its Magma.

Let us analyze how EDA tools maintain data – you have a set of base wrappers, which give rise to a little specialized wrappers and which may further be branched off into a million different highly specialized classes. Suppose, you had two different tools, a.k.a two different sets of base wrappers. Now, to be interoperable, you would have to fit in translators, to convert from one format to another.

Not too big a job – hell even one of us, underpaid developers in third world countries, could do that.

However there is a problem when you are dealing with stuff like DFM. In one short word, I would call the problem "incremental".

What happens is that whenever there is a change in the data – it may be change in netlist, or change in operating conditions, or change in naming conventions – there are two ways to do it. One could, iterate over the whole data and identify wrappers of data to-be-changed.. or … we could register what are called callbacks, which are hooks into change-procedures, and which actually change data. Maintaining callbacks between two sets of disparate data containers and keeping it consistent is very tough.

Now, let us look at US Patent #6,505,328, a patent which belongs to Magma.  The interesting part is

…

What is claimed is:

1. A common data model representing a circuit that will be fabricated on an integrated circuit chip comprising:
a data representation including a plurality of objects that together represent the circuit, certain ones of the objects including a netlist portion that represents a corresponding portion of the circuit, and each of the objects: being logically correlated to at least one other object so that all of the objects describe the circuit;

…. 

15. The model according to claim 1 wherein the area query takes place either immediately after synthesis or during final placement and routing. 

Very interesting. There is such a tight correlation between synthesis and layout data – area based queries also work. Which means that the underlying data model supports multiple layers of cascaded callbacks (where one callbacks trigger another) and consistency at several levels of design abstraction.

Why is this needed? Designers deal with changing design representations – they play around with operating conditions, timing constraints, area, power and a whole lot of things. Now add foundry data to it. And you have the need to maintain consistency and speed when the user changes a little stuff. Things should be incremental – if they need to be recomputed completely, then all the Ph.D projects to date would have been multimillion dollar products.

But of course, there could be a major data-structure redesign going on in the other places too. Only time would tell.

 del.icio.us Tags: eda software

Magma support for OpenAccess

This is not surprising, but tiresome. Magma supports the Si2 OpenAccess format for inter-operability. Now this would be a non-issue except for one thing – Magma apparently has cool internal data structures to support its entire unified flow. And what’s more, these data structures are patented.

However, I suppose this makes sense from a business perspective. Magma is’nt used across the entire design cycle. The front-end is dominated by the Big Three. And I suppose when big design firms ask you to read data in formats dumped by other tools, you’re gonna have to do it.

del.icio.us Tags: eda