Comcast (CMCSA) reported before the opening on July 27, hitting their numbers and reaffirming guidance. But a loss of basic cable subscribers rattled investors, and the stock closed down $1.33 that day, and it's lost almost another buck since then.

First, the numbers: Revenues were up 30.5% from last year to $7.71 billion and earnings came in at 19 cents a share, both on target. Guidance remains the same: 2007 revenue growth at least 11%, capital spending $5.7 billion and cable cash flow growth north of 14%.

The company added 670,000 subscribers for its Internet phone service, 2.1 million digital cable boxes and 823,000 new digital cable subscribers -- a very strong number. About 59% of subscribers now get digital video service. Comcast added 330,000 new broadband subscribers, and more than 26% of the video customers also get their cable modem service.

The only soft spots were that basic cable subscribers were down 95,000 in the quarter, and the company did not raise guidance in spite of the strong digital subscriber additions. Comcast said that they expected the decline in basic cable, but it was at the high end of Wall Street expectations. As for the guidance, I think that they are just sandbagging.

Harmonic (HLIT) reported earnings after the close on July 25. The consensus was for $72.65 in sales and 10 cents per share, and the company did $71.3 million, up 34% from last year, and 11 cents. Bookings were strong across the board, and they are now guiding for $150 million to $160 million in sales in the second half of the year, compared with consensus expectations for $146 million. They also predicted 44% to 45% pro forma gross margins in the second half, again above expectations.

Where is the strength coming from? Everywhere, as satellite, cable and telephone companies around the world all plunge into delivering video to the consumer's living room, and that should be followed by high-definition video. In Japan, consumers get 100 megabits-per-second (mbps) fiber-optic Internet service for less than $30 per month. Comcast (CMCSA) will be rolling out Data Over Cable Service Interface Specifications version 3 (DOCSIS-3) over the next two years. DOCSIS-3 cable modems can operate at 150 mbps, compared to the 12 mbps service that I now get. Verizon (VZ), with fiber to the living room, and AT&T (T), with fiber to the curb and then as much copper as it takes to the living room, have to respond to Comcast's new faster Internet offering. We are on the edge of a dramatic increase in Internet speeds, with a much broader area for connections, thanks to WiMAX.

We're also on the edge of a serious price war. Verizon's fiber optic business Internet price schedule was leaked -- only $40 a month for 10 megabits downloading and two megabits uploading, up to $350 a month for 50 megabits downloading and 10 megabits uploading. That will easily replace a couple of T-1 lines, which cost at least $500 a month each. Competition is heating up, and HLIT will be at the front of the markets, providing the top video technology to these companies.

Packeteer (PKTR) got quite a write-up in The Wall Street Journal in a July 30 article on Deep Packet Inspection (DPI), which PKTR does better than anyone in the world. DPI gives the customer unprecedented visibility into what is traveling over their network, letting them decide which packets should take priority, which can be slightly delayed or compressed without any negative impact, and which should be stopped because they are computer worms or viruses.

Paketeer's new iShaper is a "branch office in a box" that offers visibility, compression of data to save buying more hardware (while not compressing voice or video, to maintain quality, and acceleration of critical packets. The competitors are promising equivalent products, but only PKTR really has the goods.

The Journal article title and subhead were: "A Question of Priorities -- Faced with clogged networks, companies and college campuses get more sophisticated about which online material gets in first." Colleges are big users of DPI, because students are heavy users of high-bandwidth services like music, movie and video downloads that clog networks. A Ball State network engineer was quoted as saying: "I can't imagine a university in the United States without some kind of DPI technology in their network. It's just that important."

Rawlings Sporting Goods turned to Packeteer to solve their problem, too. During the 2005 holiday season, they had a barrage of complaints from customers about Internet orders taking too long to go through. Rawlings engineers couldn't figure out the cause of the problem until they tried a DPI demo system from Packeteer. It turned out that employees were eating up bandwidth by watching streaming videos. "The network was the last place I would have thought we had problems," said Richard Truex, network manager for Rawlings.

By deploying the Packeteer system, Rawlings was able to improve the flow of network traffic without having to buy more bandwidth, leading to major cost savings. Packeteer's system allowed Mr. Truex to create rules to guarantee bandwidth for important applications, like customer orders, and squeeze bandwidth for others. During peak business hours, streaming videos are nearly shut off to allow the flow of important data.

Knowing that PKTR has the best products for a serious problem does not guarantee that the stock will go up, though. But now that there are a couple of activist private equity funds involved with PKTR, I think something will happen sooner rather than later.

Energy Conversion Devices (ENER) makes nickel-metal hydride (NiMH) batteries for hybrid cars through its Cobasys joint venture with Chevron (CVX). These are great batteries -- I have them in my Toyota Camry hybrid -- but Cobasys had to work hard to reduce this technology's tendency to bleed oxygen at lower temperatures, and be susceptible to explosions at higher temperatures. Their reward for all their hard work has been to capture a large portion of the value-added in hybrids. Today, the percentage of the overall production cost of a regular car accounted for by electronics and software is about 20% (far more than the cost of the steel, incidentally), but 50% for a hybrid.

A new technology, nanophosphate lithium-ion batteries, is now on the market in the DeWalt portable saw. It takes only five minutes to recharge, does not emit oxygen at any temperature, and is unlikely to explode. For now, it is only applicable to smaller applications like portable power tools, but I think it will eventually scale up to hybrid car size. Combine that with the next generation of plug-in hybrids that can be charged by solar cells on your garage roof, and things could get interesting. Plugging into a service station outlet for five minutes isn't any more onerous or time-consuming than filling up at the pump.

Don't worry about these batteries threatening ENER and Chevron, because the M1 battery in the DeWalt saw is made by a company called A123 Systems and... Cobasys. ENER and Chevron are making sure that if anyone obsoletes their current products, it will be them.

At the end of July, I attended the San Francisco Money Show. While there, I sat down and talked with Gary Alexander about my thoughts on technologies in the medical field that are helping to extend our lives. Plus, I discussed a couple companies making great strides in this industry. Check out the interview here.

Semiconductor prices are a little softer than what I was looking for, so even though unit volumes are good, I am going to slightly reduce my dollar sales forecast for the year. The weakness is in memory chips, especially DRAM (Micron) and NAND flash (SanDisk) -- commodity chip producers. There are plenty of reasons to think DRAM and NAND flash prices could stabilize or even go up a bit, but it hasn't happened yet. Year-to-date total semiconductor sales through June were $120.9 billion, up only 2.1% from last year. January was up 10.8% over January 2006, but the year-to-date comparison has been declining ever since.

I had thought semiconductor sales would be up 7% this year, but I think it is prudent to trim that back to +5% until I see memory prices firming, which will probably happen when the demand for PCs and consumer electronics picks up. That would get us to $260 billion for the year instead of $265 billion. I know that seems like a trivial change, but it is the first downward revision of a semiconductor forecast that I have had to make in more than three years. It could be a straw in the wind, especially if high gas prices or credit problems inhibit holiday spending.

Natural gas prices plunged more than 10% last Monday after Hurricane Dean turned away from the U.S., the biggest one-day drop in four years. Then last Tuesday, prices fell another 3.2% after Dean weakened to a Category II storm, and September contracts hit $5.85. But prices came right back on Wednesday to climb back above $5.85, which held today. This sets up a good-sized rebound move at least to the $6.50 breakdown point.

The move in gas matches a corrective pattern in crude oil that started at the end of July and has brought prices down from $78 a barrel to just under $70. This is the same consolidation range that we saw in mid-June, before the breakout to $78, and I expect it to end the same way. These full tests down to the last breakout level shake out as many bulls as possible, before the trend resumes. In this case, summer driving season is coming to a close and there haven't been any serious hurricanes, yet oil prices are still near $70 as we enter the heavy industrial use period in the fall. With China growing well above that government's 10% target and world production flat to falling, it isn't hard to see why we keep getting higher lows and higher highs in oil prices.

Interested in robots? I am. I've been following robotics companies for over 30 years. But this is a tough business to be. Many early companies have gotten their heads handed to them, but I think the revolution in consumer robotics is just getting underway. And to me that spells opportunity -- if you know which companies to look at, that is. One of my favorites is actually pretty well known for its house cleaning robots. You know those little, flat, round vacuums that zoom around your living room picking up dirt, without you having to push it around? The company that produces these nifty tech gifts is iRobot (IRBT).

The company, though, isn't limited to just neat consumer gadgets. It is also producing robots for our military. A week or so ago, The Associated Press broke a story that: "The Pentagon plans to purchase up to 3,000 additional robots to be used by U.S. soldiers in Iraq and Afghanistan to detect explosive devices and roadside bombs." It's a little more complicated than that, but this would be a big contract for the company, worth around $180 million based on iRobot's selling price of $60,000 per PackBot. That's almost as much as the company's entire revenues in 2006. One thousand PackBots will be ordered this year, with 2,000 more to follow over the next five years. (Did you know that the Department of Defense is planning to be in Iraq and Afghanistan until 2012?)

One thousand robots, or $60 million, is still a big order, but iRobot may have to split it. Foster Miller, a subsidiary of Britain's QinetiQ Group, makes the Talon bomb disposal robot with similar capabilities at about the same price. There's even a chance Exponent's MARCbot could get some of the business, even though it is not as advanced as the PackBot, as it is cheaper. Still, even sharing the contract will generate a nice chunk of change.

One company that is not likely to get a share of this contract is Robotic FX, formed by an ex-iRobot employee. They make a tactical surveillance robot called Negotiator that looks like the PackBot, but they claim will sell for only $20,000. iRobot has sued them in Federal and state court for patent infringement and misuse of confidential information.

Another threat to iRobot that I am watching is the new armed robot from QinetiQ's Foster Miller. Only three of these bad boys have been deployed so far, but they are designed to wheel into combat or door-to-door searches and shoot people. According to National Defense magazine, the U.S. Army started putting Foster Miller's SWORD armed robots into combat a couple of months ago, and it can get rid of roadside bombs in addition to (or possibly at the same time as) shooting people.

iRobot does not make an armed robot, and they've never talked about a development program. The Army says that reaction by soldiers has been so positive that they want 20 more SWORD robots immediately, part of an 80-unit purchase authorization. According to the Pentagon, robots will be 33% of our combat forces by 2015, as part of the $127 billion Future Combat System Program. I suspect this is a direction that iRobot has to go in order to keep up with competition.